research paper on multimedia communication

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Sensors (Basel)

Internet of Multimedia Things (IoMT): Opportunities, Challenges and Solutions

Yousaf bin zikria.

1 Department of Information and Communication Engineering, Yeungnam University, 280 Daehak-Ro, Gyeongsan, Gyeongbuk 38541, Korea; moc.liamg@airkiznibfasuoy

Muhammad Khalil Afzal

2 COMSATS University Islamabad, Wah Campus, Wah Cantt 47010, Pakistan; kp.ude.hawtiic@lazfalilahk

Sung Won Kim

With the immersive growth of the Internet of Things (IoT) and real-time adaptability, quality of life for people is improving. IoT applications are diverse in nature and one crucial aspect of it is multimedia sensors and devices. These IoT multimedia devices form the Internet of Multimedia Things (IoMT). It generates a massive volume of data with different characteristics and requirements than the IoT. The real-time deployment scenarios vary from smart traffic monitoring to smart hospitals. Hence, Timely delivery of IoMT data and decision making is critical as it directly involves the safety of human beings. In this paper, we present a brief overview of IoMT and future research directions. Afterward, we provide an overview of the accepted articles in our special issue on the IoMT: Opportunities, Challenges, and Solutions.

1. Introduction

Internet of Things (IoT) devices has limited memory and processing capabilities [ 1 ]. Hence, these constrained devices rely upon efficient routing protocols [ 2 ] and standardized communication stack [ 3 ]. Technological advancements in 5G [ 4 ], intelligent 5G-based IoT [ 5 ], IoT operating systems (OS) [ 6 ], data-driven intelligence in wireless networks [ 7 ], scheduling approaches for heterogeneous content-centric IoT [ 8 ], congestion avoidance techniques in IoT using data science [ 9 ], vehicular ad hoc networks (VANETS) [ 10 ], Information-centric networks (ICN) [ 11 ], reinforcement learning-based solutions for next-generation networks [ 12 , 13 ], coexistence networks [ 14 ], IoT adaptation in agriculture [ 15 ] and Healthcare IoT [ 16 ] are helping to realize to connect everything and anywhere.

Internet of Multimedia Things (IoMT) devices are different from IoT devices. It requires bigger memory, higher computational power, and more power-hungry with higher bandwidth [ 17 ]. Figure 1 shows the key data characteristics of IoT and IoMT. The real-time deployment scenarios vary from industrial IoT, Smart cities, Smart hospitals, smart grid, smart agriculture, and smart homes. The main characteristic of IoMT is the timely and reliable delivery of the data. Therefore, it imposes strict quality of service (QoS) requirements and demands efficient network architecture. The users perspective of QoS is known as quality of experience (QoE). QoE can be further characterized as objective or subjective. The users Objective QoE is challenging to measure and dramatically varies according to the needs. However, service providers concern with the subjective QoE to evaluate the network mean opinion score (MOS). The multimedia data is increasing multifold. It raises new challenges to transmit, process, store and share the data. Processing requires new techniques for edge, fog and cloud devices. Further compression and decompression techniques are introduced for the storage of multimedia data. Routing protocol for low-power and lossy networks (RPL) is the standard IoT routing protocol. It needs further development by considering energy-aware, load balancing, fault tolerance, and delay aware IoMT deployment scenarios.

Key Data Characteristics of IoT and IoMT.

IoT characteristics support multimedia communications; however, multimedia applications are bandwidth-hungry and delay-sensitive. The rapid growth of multimedia traffic in IoT has led the way to innovating new techniques to meet its requirements. IoMT devices require higher bandwidth, bigger memory, and faster computational resources to process data. Typical communications include multipoint-to-point and multipoint-to-multipoint scenarios. Real-world multimedia applications include emergency response systems, traffic monitoring, crime inspection, smart cities, smart homes, smart hospitals, smart agriculture, surveillance systems, Internet of bodies (IoB), and Industrial IoT (IIoT). Dynamic networks, heterogeneous devices and data, strict QoS, and delay sensitivity and reliability requirements over resource-constrained IoMT pose humongous challenges for multimedia communication in IoT. Network-on-chip architecture [ 18 , 19 ] is one of the viable solution to improve the user quality of experience. Figure 2 shows versatile IoMT applications.

IoMT Applications.

The rest of the paper is organized as follows. Section 2 provides future research directions. Section 3 summarizes the accepted paper. Finally, Section 4 concludes the paper.

2. Future Research Directions

Molecular communication exploits the transmission and reception of information encoded in molecules. Molecular communications have the potential of becoming the main technology for the execution of advanced medical solutions. The key research challenges in the molecular communication are the interoperability between molecular communication and the other networks, energy-efficient models and protocols for molecular communication, and implantation of reliability in the molecular communication.

Billions of resource constraint devices will be connected in the IoT. The available spectrum is far from enough to support IoT communication systems. Optimal resource allocation for critical multimedia traffic is a key challenge for IoMT. The use of artificial intelligence (machine learning, deep learning) can improve the energy-efficient resource allocation in IoMT.

Device-to-device (D2D) communication in LTE-A will establish direct communication with the device in its communication range. Potential advantages of D2D communication are increased network spectral efficiency, energy efficiency, reduced transmission delay, traffic offloaded base station, and less congestion in the cellular core network. IoMT can take the benefits of the advantages provided by D2D communication. Interference, radio resource allocation, power control, and QoE improvement for cellular users are the key research areas in D2D communication for IoMT.

Energy Efficient Operation and Protocols are the key requirement of IoMT. Many multimedia traffic sources in IoMT may rely on battery-powered sources with limited energy and/or may not be easily accessible for recharging purposes. Similar to WSNs/IoT, the energy-efficient operation, protocols design (i.e., medium access control and routing protocols), and the need to optimize the network lifetime remains a critical challenge for IoMT. Based on the specific application and deployment environment, energy-efficient protocols can be designed for IoMT.

The Internet of Multimedia Nano-Things (IoMNT) is defined as the interconnection of multimedia nano-devices with communication networks and the Internet. The potential applications of IoMNT are security, biomedical, defense, and industry. The main research challenges in IoMNT includes novel medium access control techniques, addressing schemes, neighbor discovery and routing schemes, QoS-aware cross-layer communication module and security solutions for the IoMNT.

Multimedia-oriented IoT over vehicular networks is increasing drastically. Today, vehicles have the capability of supporting real-time acquisition and transmission of the multimedia traffic generated by the built-in IoT devices. However, due to high mobility, density, and random wireless channel conditions, the performance of the delivery of multimedia contents significantly reduces in vehicular networks. Rate adaptation, multimedia delivery over heterogeneous devices, robust video encoding, scalable, and timely delivery of multimedia contents are the key research challenges in IoMT over vehicular networks.

3. A Brief Review of Articles of This Special Issue

The immense growth in multimedia traffic over the scarce licensed cellular spectrum has inspired to use unlicensed spectrum below 6 GHz for Long Term Evolution (LTE). However, Wi-Fi uses the same unlicensed band, and this gives rise to the issue of coexistence and fairness of two different technologies in the context of physical and link layer protocols. LTE in the Unlicensed (LTE-U) and LTE License Assisted Access (LTE-LAA) has been proposed in the literature for IoT system. The Third Generation Partnership (3GPP) has standardized LAA for industrial IoT. The coexistence mechanism of LAA follows Listen Before Talk (LBT), which is the same process of Wi-Fi system coexistence i.e., Carrier Sense Multiple Access (CSMA). LTE-U operates a carrier ON/OFF switch policy in duty cycles to maintain fairness in LTE and Wi-Fi transmissions. This mechanism causes spectrum inefficiency. Bajracharya et al. [ 20 ] proposed a Machine Learning (ML)-based Adaptive Duty Cycle (ADC) and Dynamic Channel Switch (DCS) mechanism for network to access channel in dynamic network scenarios. ADC and DCS exploit Q-learning to determine the best policy to select an optimal channel and duty cycles. ADC reserve a specific number of sub-frames for Wi-Fi, whereas DCS avoids congested channels for LTE-U users. Performance evaluations are presented in comparison with the fixed duty cycle and channel occupancy time approach. Results show that their proposed method outperforms other methods in the context of fairness and throughput.

With the exponential growth of the IoT, the interaction of multiple physical devices is of extreme importance. These devices are often integrated using Radio Frequency Identification (RFID). The RFID automatically recognizes the object details by reading the physical objects. The system reader, which is equipped with a backend server, uses radio frequencies to communicate with the objects with RFID tags. It makes the practical usage of RFID very vast. Security is the most vital aspect of communication for authentication and securing private data. RFID-based security is beneficial in multiple ways, as RFID does not require a light source and line of sight scenario for communication. Hence RFID can be deployed to sensor monitoring, access control, real-time inventory, and security-aware management systems. However, due to limited computational and memory resources on an RFID tag, limited cryptographic operations can be applied. Therefore, an eavesdropper can forge and access the user’s private data. David et al. in [ 21 ] proposed a hash-based RFID authentication mechanism for Context-Aware Sensor Management System (CASMS) to provide security to prevent attacks such as replay, man-in-the-middle, and desynchronization. Hash-based RFID authentication is the five-phase mechanism, namely pre-phase registration, reader pro-tag request and response, tag mutual session key authentication, back end server key authentication, and session key updating. Performance analysis is made based on the Packet Delivery Ratio (PDR) and End-to-End Delay (E2E). Results depict that the proposed model significantly improves PDR and E2E.

Water is the soul of life and essential to the well-being of every person, economy, and the ecosystem on Earth. More than 70% of the Earth is surface is covered by oceans, and they are critical to maintaining the weather and temperature around the globe and providing a means of transportation. However, more than 90% of oceans are unexplored even to the extent that they are still unseen by humans. IoT paved the way to explore and collect the data by connecting different types of networks underwater. Such networks are known as the Internet of Underwater Technology (IoUT) is an emerging technology to support Underwater Sensors Networks (UWSNs) for exploration of undiscovered marine resources. UWSN using communication cables and sensors and maintenance cost is very high. Therefore, underwater wireless communication is proposed. However, UWSN wireless communication is challenging due to environment and propagation losses, which include high noise, Doppler spreading, path loss, multi-path signal propagation, and high power consumption. To overcome these issues, Faheem et al. in [ 22 ] proposed cross layered QoS Aware routing Protocol (QoSRP). The proposed scheme is composed of underwater channel detection, channel assignment, and packets forwarding. The QoSRP selects detects the high probability vacant channel and assigns the high data rate channels to an acoustic sensor node. The QoSRP also balances the traffic, avoids congestion, and data path loops to increase PDR and throughput of the system with minimum delay along the path.

Ultra Wide band (UWB) features include higher bandwidth, and it is one of the viable technologies for IoMT applications. The integration of UWB in health critical IoT applications can provide an effective and reliable solution for the monitoring of patients. Ataxia patients suffer from abnormal movement, and that severely affects walking activities. The walking activities can be classified as a normal walk, difficulty walking in a straight line, walking with heavy steps, and forward bending walking. All of these walking patterns except normal walking shows abnormality. Zilani et al. [ 23 ] proposed a scheme to cater to this problem. They set up the testbed in an indoor environment. They collected sample walk patterns and classified it using Support Vector Machine (SVM) algorithms, namely SVM-based Sigmoid Kernel Function (SKF) and Radial Basis Function (RBF). Results show that RBF performs better than SKF. The drawback of the proposed scheme is that it is only tested for a single person. Hence, it should be extended to monitor multiple persons.

Security, privacy, and trust remain challenges in IoMT because of the openness and heterogeneity of IoMT. Access control is used to protect the confidentiality and integrity of constrained resources in IoMT services. It provides a solution to avoid any unauthorized access for multimedia applications in IoT services. However, due to increasing the number of users and multimedia services offered by the IoT platform, the access control system is becoming more and more multifaceted. Besides, access control policy evaluation reduces the performance of IoMT applications. Therefore, Meiping Liu et al. [ 24 ] proposed an Attribute-Based Access Control (ABAC) policy retrieval method to improve the performance of access control policy evaluation in multimedia networks. To rebuild the policy decision tree, an attribute value level, and the depth index is introduced, thus, improve policy retrieval efficiency. Policy analysis is performed with a different number of rules and the increasing complexity of the policy. Results indicate that the proposed method is more efficient and scalable than the existing access control schemes.

Increasing demand for data-intensive applications is growing users’ data requirements exponentially. However, spectrum scarcity is the biggest hurdle in meeting users QoS. One of the viable solutions is to reuse and share the spectrum among the users to fulfill users demands without compromising the user’s experience. Even though unlicensed spectrums are available for free, they are already overcrowded. Different network technologies such as 5G and Wi-Fi use different spectrum access mechanisms, and sharing the spectrum among them is a trivial task. To allow fair coexistence between 5G and Wi-Fi networks operating in the same spectrum, LBT is introduced to work in parallel with the CSMA for channel access. RL techniques can be adapted to make a spectrum access mechanism to learn network conditions itself and adapt to the network changes accordingly. Consequently, the network becomes sustainable and self-adaptive. Neto et al. in [ 25 ] proposed Q-Learning to LTE-U to adjust the duty cycle parameters to reduce coexistence interference and improve the system data rate. A saturated network scenario is considered to evaluate the proposed scheme in the ns-3 simulator. The proposed algorithm performs well in a multi-cell coexistence network scenario. Hence, it improves overall system performance by achieving a higher data rate for users and systems compared to the existing conventional mechanism.

4. Conclusions

Six papers in this SI presented state-of-the-art research trend in the area of IoMT opportunities, challenges, and solutions. The papers presented an interesting discussion and novel ideas for the readers. The guest editors would like to show appreciation to the authors and thank all the anonymous reviewers for providing constructive feedback to improve the overall quality of all the accepted papers. We would also like to thank sensors Editor in Chief Prof. Dr. Vittorio M.N. Passaro, Associate Editor in Chief of the IoT section Prof. Dr. Raffaele Bruno, and managing editor Missy Wu for the invaluable help and productive advice in finalizing this SI.

Acknowledgments

This research was supported in part by the Brain Korea 21 Plus Program (No. 22A20130012814) funded by the National Research Foundation of Korea (NRF), in part by the MSIT(Ministry of Science and ICT), Korea, under the ITRC(Information Technology Research Center) support program(IITP-2019-2016-0-00313) supervised by the IITP(Institute for Information & communications Technology Planning & Evaluation), and in part by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (2018R1D1A1A09082266).

Abbreviations

The following abbreviations are used in this manuscript:

Author Contributions

Conceptualization, Y.B.Z. and M.K.A.; Writing—Original Draft Preparation, Y.B.Z.; Writing—Review & Editing, Y.B.Z., M.K.A., S.W.K.; Supervision, S.W.K. All authors have read and agreed to the published version of the manuscript.

This research received no external funding.

Conflicts of Interest

The authors declare no conflict of interest.

• Open access
• Published: 15 September 2020

Smart multimedia learning of ICT: role and impact on language learners’ writing fluency— YouTube online English learning resources as an example

• Azzam Alobaid   ORCID: orcid.org/0000-0002-2990-6987 1  

Smart Learning Environments volume  7 , Article number:  24 ( 2020 ) Cite this article

31k Accesses

22 Citations

3 Altmetric

Metrics details

This work seeks to determine if and how much the smart learning environment of Information and Communications Technology (ICT) tools like YouTube can help improve learners’ fluency of language use and expression in their daily written communication. This research highlights and takes advantage of the potential role and features of multimedia brought to the language learner by the ICT tools, taking YouTube online English learning resources as an example of this smart learning environment. This work hypothesizes that learners who engage with, expose themselves more to and leverage such online language materials could develop their fluency of daily language use and expression in writing over time. The findings of this research show that there is a statistically significant difference in some but not all aspects of the learners’ writing fluency; basically, the accuracy and organization of ideas as qualitative dimensions of fluency improved after the actual exposure to YouTube over five months as long as factors like engagement, enhancement and intelligibility are provided by its multi-mediated input. However, other aspects of fluency in writing could develop slightly but with no statistically significant difference. Also, compared to other sources of language learning in the learners’ environment, multimedia educational tools developed by ICT like the widely known platform YouTube can be more effective and thus strongly recommended equally for language learners and teachers where optimization of writing fluency is the target of learning. This paper is a work-in-progress that investigates the role and impact of smart learning environment of ICT multi-media on English language learners’ fluency and accuracy of use and expression in speaking and writing.

Introduction

ICT in the field of education is the integration of various technologies of information and communication so as to leverage its capacity for the optimisation, enhancement and creating of a better learning environment and smoother learning process. A wealth of research in the literature showed the significant and positive impact on students’ achievements through the increase in the use and exposure to ICT in education. Aoki ( 2010 ) in a report findings by the National Institute of Multimedia Education in Japan showed that “the students with continuous exposure to ICT technology through education demonstrated better ‘knowledge’, presentation skills, innovative capabilities, and were ready to take more efforts into learning as compared to their counterparts”. Moreover, ICT has impacted multimedia learning of language with its attractive and interactive strengths to provide easy to reach multimedia language materials (MMLM); such MMLM are packaged with graphical, textual, animated, audio and video materials delivered to the end-user through wide variety of electronic devices, primarily via computers, smart boards and phones. ICT, whose implementation is intended to create smart learning environment— YouTube online English learning resources as an example of this smart learning environment, is synonymous with smart learning environments in this work. However, it would be useful to define smart learning environments although there is no completely agreed upon definition for them, yet they are often understood as an improvement of physical environments with novel technologies to provide a smart, interactive classroom with increased interactivity, personalized learning, efficient classroom management, and better student monitoring (Yesner, 2012 , as cited in Libbrecht, Müller, & Rebholz, 2015 ). Smart learning environments are related to ambient technologies, describing learning environments, which exploit new technologies and approaches, such as ubiquitous and mobile learning, to support people in their daily lives in a proactive yet unobtrusive way (Buchem and Pérez-Sanagustín, 2013 ; Mikulecký 2012 , as cited in Libbrecht et al., 2015 ). The rational and significance for integrating ICT tools for creating smart learning environments in foreign/second language education lies in the fact that ICT multi-faceted features in the domain of language learning are many, especially those that can be beneficial for language learners like personalization, networking and interactivity, inclusiveness, richness of authentic and engaging input.

Personalization is one major facilitative characteristic of modern ICT in education also known as customization or individuality of choice of materials where a service or a product can be tailored to cater for specific individuals’ or group’s needs. Put differently, each student can now learn at his/her own pace and space and instructors can accommodate and cater for the individual needs and interests of learners. This feature can be hypothesized to enhance multi-media learning through the adjustment of a given language input that learners may be able to get more and clearer input and thus greater or more fluent output. More on personalization and the integrity of knowledge, Vaughan ( 1993 ) suggested that when the user or viewer of the project can adapt and control what and when these elements are delivered, it is interactive multimedia; more usefully, this interactive multimedia has become hypermedia providing learners with a structure of linked elements where now materials can be widely navigated, interacted with and exchanged. In this research ICT multimedia is employed as both interactive and hypermedia catalyst where a “structure of linked elements of knowledge” is bringing an array of MMLM to learners through their devices. Also, personalization includes the use of YouTube closed captions and its adjustable settings related to font size, color, opacity and the playback speed for learning and improving writing.

Networking and interactivity are the use of social networks in both private and academic life, for example YouTube. (Mukhaini, Al-Qayoudhi, & Al-Badi, 2014 ) “These are tools used to enable users for social interaction. The use of social networks (SNs) complements and enhances teaching in traditional classrooms. For instance, YouTube, Facebook, wikis, and blogs provide a huge amount of material on a wide range of subjects. Students can, therefore, turn to any of these tools for further explanations or clarifications”. This feature can be hypothesized to enhance the reception and production of input giving leaners a chance for language practice and active exposure.

Inclusiveness means that a wider diversity of people can make (easy) use of it. Rice ( 2011 ) Inherent in inclusive education is the notion that reform and improvements should not only focus on children with disabilities but on “whole school improvement in order to remove barriers that prevent learning for all students” (GeSCI, 2007 , as cited in Rice, 2011 ). Inclusive schools can “accommodate all children regardless of their physical, intellectual, emotional, social, linguistic or other conditions” (Article 3, Salamanca Framework for Action). However, inclusive education is not a synonym for special needs education or integration techniques but an “an on-going process in an ever-evolving education system, focusing on those currently excluded from accessing education, as well as those who are in school but not learning” (UNESCO 2008 ). According to Rice ( 2011 ) “There is a wide variety of accessible ICTs currently available which can help overcome reduced functional capacity and enable communication, cognition and access to computers.” This feature can mean welcoming and accommodating more learners through the potential outreaching and connecting with everyone.

Rich provider for authentic language input, context and culture as in the easy-to-get authentic input that is quantity- and quality-wise useful. Grzeszczyk ( 2016 ) noted that “CALL was expanding and introducing tools helped teachers to supply learners with more up-to-date, authentic, target-specific, and learner-oriented materials”. According to Pun ( 2013 ) “using multimedia technology offers the students with more information than textbooks and helps them to be familiar with cultural backgrounds and real-life language materials, which can attract the students to learning”. More on culture, Pun ( 2013 ) observed that “the learners, through the use of multimedia technology, will not only improve their listening ability, but also learn the culture of the target language; this brings about an information sharing opportunity among students and makes them actively participate in the class activities that help the students to learn the language more quickly and effectively”. In terms of context, “the utilization of multimedia technology can create a context for the exchange of information among students and between teachers and students, emphasizing student engagement in authentic, meaningful interaction” (Warschauer & Meskill, 2000 , as cited in Pun, 2013 ) as in “using multimedia in English teaching/ learning can be an appropriate method to help students to get a sense of the sociocultural context in which the language is used” (Kramsch, 1999 , as cited in Yang, 2008 ), “as well as raising students’ language awareness under the framework of World Englishes” (Kunschak, 2004 , McKay, 2002 , as cited in Yang, 2008 ).

ICT can provide a whole raft of engaging multi-media materials as learners can be all the more free to choose and adjust their learning materials. A wealth of research has shown that “using social media as an educational tool can lead to increased student engagement motivating them to learn more (about) languages by way of audio, visual and animation support” Tarantino, McDonough, and Hua ( 2013 ). (See also Annetta, Minogue, Holmes, & Cheng, 2009 ; Chen, Lambert, & Guidry, 2010 ; Junco 2012 ; Patera, Draper, & Naef, 2008 ). (Izquierdo, Simard, & Garza, 2015 ) confirmed previously established research in education that ICT makes it easier for learners to access language materials, stressing an existing correlation between second language learning and the use of multi-media materials in a computer-enhanced language learning milieu showing an impact on learning behaviour with increased motivation. This said, ICT multi-media could be enriching for the learners’ language experience as in helping learners write fluently (quantity- as well as quality-wise). This research underscores that the language used in this input can be taken as a model for the language learner. Thus, in addition to other previous and coming input possibly acquired by the learner, the new input could be in effect increasingly adding up to the learner’s knowledge of language and developing their output to be more fluent.

With the above-mentioned plausible advantages of ICT in mind and that modern ICT technology like YouTube has impacted multimedia in education as far as it can be seen through a wide range of educational channels, it’s worth questioning this impact from a desirable angle by language learners and instructors, which is the fluency of writing performance.

A far-reaching definition of fluency in writing that would suit the purpose of this study is based on a number of perspectives of fluency in the literature combined in a study by (Atasoy & Temizkan, 2016 ) “the act of writing the maximum number of language units in a short period of time while also paying attention to accuracy, the coherent and consistent organization of ideas within the text, and the usage of words and sentences in a complex manner.”

The focal point of this study is that, can learners of English make a progress in a better (in the sense of quality and quantity) communicative writing as long as the seemingly learning environment by way of multi-media technology of ICT cannot be more helpful as it is as facilitative as it sounds (both in terms of technical capabilities and language materials and content information possibilities) for learners of English in writing?

This study attempts to reach an understanding of what and how much, if any, can be developed of learners’ written performance in terms of fluency given that they are to varying degrees exposed to and engaging with ICT multi-media during their learning process.

RQ. Can exposure to and engagement with ICT educational multimedia like YouTube have some effect on the development of learners’ fluency of language use and expression in writing?

Literature review

This research examines the multimedia role of ICT in language learning as input provider of language materials, assuming that ICT YouTube technology as input provider and enhancer, and its impact on learners’ development of fluency in writing. Therefore, it would be necessary to look into the relevant input theories of language learning and those of multimedia trying to connect the dots between them; the theories adopted in this research are the language learning theories of Comprehensible Input, Input and Interaction, Comprehensible Output, Input Enhancement, Noticing the Gap and the Multimedia Learning Theory.

The first and foremost stage of language acquisition assumed by Krashen and his proponents is the introduction of “comprehensible input” which is defined as “a language input which can be comprehended by listeners albeit not understanding all the words and structures in it. This input is described as one level above that of the learners if it can only just be understood” (British Council, 2020a , b ). ‘Comprehensible input’ is the crucial and necessary ingredient for the acquisition of language supplied in a low anxiety situation, containing messages that students really want to hear. However essential to the language development, “comprehensible input is held to be a necessary, though not sufficient, condition for SLA” (Krashen, 1985 ; Long, 1983 ). Long’s Input and Interaction Hypothesis (Long, 1985 ) argues for the significance of both input comprehension and modifications in order to facilitate language acquisition, i.e., through negotiated interaction of discourse structure and modified input. Following Krashen’s comprehensible input hypothesis (1992, 1994), “Multi-media Instruction (MI) research has provided learners with rich exposure to the L2 in meaning-based tasks built upon different media features” (Plass & Jones, 2005 , as cited in Izquierdo et al., 2015 ).

According to Swain ( 2005 ), the output hypothesis “asserts that the act of producing language (speaking or writing) constitutes under certain circumstances, part of the process of second language learning” (Swain, 2005 , p. 471, as cited in Pannell, Partsch, & Fullver, 2017 ) and “that even without implicit or explicit feedback provided from an interlocuter about the learners’ output, learners may still, on occasion, notice a gap in their own knowledge when they encounter a problem in trying to produce the L2” (Swain & Lapkin, 1995 ).

Input enhancement, Smith ( 1991 ) examines the concept of “‘consciousness raising’ in second language learning, i.e., how certain features of language input become salient to learners suggesting different ways for making input salient and different ways in which such salience may affect the learner’s knowledge and performance in language learning”. Relating to the concept of consciousness raising is what is now termed in neurolinguistics as metacognitive awareness. Just as it is essential to effective learning, metacognition is an important part of successful technology adoption (Gurbin, 2015 ).

The noticing hypothesis (Schmidt, 1990 , 2001 , as cited in Schmidt, 2010 ) argues that “input does not become intake for language learning unless it is noticed, that is, consciously registered”.

Mayer ( 2009 ) defines a “multimedia environment as one in which material is presented in more than one format – such as in words and pictures” where according to his “multimedia principle”, Mayer ( 2002 ) premises that “people can learn more deeply when they receive an explanation in words and pictures rather than words alone”. In the same vein, “according to the sensory modalities view, in his perspective, multimedia means that two or more sensory systems in the learner are involved. It focuses on the sensory receptors the learner uses to perceive the incoming material – such as the eyes and the ears”. In this respect, the signalling principle of multimedia states that students engage in “deeper learning when key steps in the narration are signalled rather than non-signalled” (Mayer, 2005 ). Signals give cues to the learners about what words and pictures to notice and enables their organisation. This means that linguistic elements need to be linked to some visual stimuli so as to assist learners’ storage of the new linguistics elements in their long-term memory (Matus, 2018 ).

In relation to these multimedia theories, a number of multimedia technology studies were conducted in the field of language learning focusing on different language aspects and different multimedia applications and their potential impacts and facilitative features. For example, the multi-media effect of captions or subtitles inclusion whilst watching films dealt with issues like the listening comprehension improvement and vocabulary acquisition. (Yoshino, Kano, & Akahori, 2000 ) examined the listening comprehension of Japanese EFL students and found that foreign subtitles inclusion can be “helpful but native-language subtitles provide no benefit or less benefit” in this regard. (Mitterer & McQueen, 2009 ) supported the previous studies that second language learners can boost their listening ability when watching films with “foreign-language subtitles as this can improve repetition of both previously heard and new words, the latter demonstrating lexically-guided retuning of perception. While native-language subtitles can help only recognition of previously heard words but harm recognition of new words”. In terms of vocabulary acquisition, in their study (Winke, Gass, & Sydorenko, 2010 ) mentioned “a number of observations about the use of captions, confirming previous research that captions are beneficial because they result in greater depth of processing by focusing attention, reinforce the acquisition of vocabulary through multiple modalities, and allow learners to determine meaning through the unpacking of language chunks”. Xiao ( 2007 ) conducted a study on the effect of the use or interaction with native speakers through video conferencing as a multimedia for oral practice of speaking in English on learners’ oral improvement of fluency, accuracy and complexity. He found “a significant improvement in fluency, a slightly significant improvement in accuracy, but no improvement in complexity for the L2 learners” as a result of this kind of exposure to English to interact with native speakers.

The potential of using ICT multimedia with its features of personalization, networking, inclusiveness, richness of engaging input, context and culture like YouTube for language learning and its impact on other aspects like the fluency of use and expression of writing, however, has not been sufficiently investigated. This research is mainly interested in the role of the foreign language subtitled multimedia (like YouTube) effect on learner’ improvement of writing fluency. In this research, “multimedia” means the availability of both speech and text as implied by the sensory modalities view of multimedia contrasted with mono-media as having either of them, i.e., speech/ text, Mayer ( 2009 ).

To connect the dots, this work proposes the use of ICT as a multimedia source of comprehensible input which is hypothesized to lend itself to help learners producing comprehensible output, i.e., or more fluent use and expression of language in writing; this is hypothesized to be achieved first, as ‘learners consciously by themselves or made conscious by the teacher to notice their gap(s), (especially in terms of gaps related to accuracy of language use), in the provided input—what becomes intake for learning’ as implied in the Noticing the Gap Hypothesis; second, through ‘negotiated interaction (with the teacher or other learners using some given multi-media material in question) and modified input’ as implied by Long’s Input and Interaction Hypothesis (Long, 1985 ). One effective way to let input noticing and negotiating happen is through the use of the ICT personalization and networking and interactivity features, respectively; “the process by which language input becomes salient to learners” as indicated by the Input Enhancement Hypothesis. These workings entailed from these language learning theories collectively can arguably be practically put into action within the Sensory Modalities View of Multimedia framework when the sensory receptors of the eyes and the ears of the learner are used to perceive the incoming material of text and speech respectively in the ICT leaning environment.

Our multimedia approach takes to the full a great advantage of the multiple personalization features provided by the ICT technology available in this research example, i.e., YouTube, to make a given input as much salient or comprehensible and reproduceable as needed; such features are mainly the optional text aligned with the speech, with care given to correct spelling, placement of punctuation marks and capitalization; font size, colour and opacity; and the playback speed. Such features are considered as highly significant and facilitative for both comprehending the meaning and noticing the form(s) of the language presented in a given episode on such YouTube channels. Familiarizing learners with or controlling these features in proportion to the learners’ needs can perhaps better benefit learners navigate their learning of the language in hand and make it more accurate and fluent; that is to say, understand more and faster meaning and notice more and clearer forms, so much so that their knowledge of the language forms would not only be increasingly informed and enriched but also the number of language errors be it grammatical (omission, addition, mis-ordering), lexical (misinformation, misspelling, informality) or mechanical (punctuation, capitalization) would be reduced (for the categorization of error types and analysis used in this research this work refers to the taxonomy by (Dulay, H. et al., 1982 , as cited in Ellis & Barkhuizen, 2005 ) and the work of Ferdouse ( 2013 ). From the Input and Interaction view such learning is supposed to happen due to the learners being exposed to and engaged with this multi-mediated authentic input which can be set as a model to acquire or learn language from.

Having the plausible advantages of ICT in mind supported by research in the literature, this work attempts to first activate and encourage the use of YouTube as an example of ICT in class as well as out of class and then check on its likely impact to find answers to questions like what and how much linguistic input, if any, learners get through ICT as far as fluency in communicative writing in English is concerned. Our approach explores, in terms of exposure time range and engagement degree, the various potential language learning resources in the learning environment of this sample group, including the ICT YouTube as one of these sources; then, we correlate these sources, which are deemed as explanatory variables in this study, with the learners’ actual performance of communicative writing fluency. This work suggests the BBC Six-Minute English YouTube Channel as a case study.

This study selected and relied on exposure time range and engagement rate as factors or indicators of learning due to their huge popularity in literature as significantly influential drivers of learning foreign/ second languages (For the significance of exposure time see Benson, 2001 ; Ellis, 2002 ; Krashen, Long, & Scarcella, 1979 ; Peregoy & Boyle, 2005 . For the significance of engagement see Astin, 1984 , 1993 ; Benek-Rivera & Matthews, 2004 ; Bertin, Grave, & Narcy-Combes, 2010 ; Sarason & Banbury, 2004 ). Early studies defined student engagement primarily by observable behaviors such as participation and time on task (Brophy, 1983 ; Natriello, 1984 ). Researchers have also incorporated emotional or affective aspects into their conceptualization of engagement (Connell, 1990 ; Finn, 1989 ). It can be understood from these definitions that engagement include feelings of belonging, enjoyment, and attachment, which is how this study defined the engagement for its analysis. Time range of exposure was defined for this study as “the contact that the learner has with the language that they are trying to learn, either generally or with specific language points”. (British Council, 2020a , 2020b ).

Research methods

This research is a longitudinal study investigated patterns within time-series data. The performance of a single group of participants was measured both before and after the experimental treatment. It was conducted over a period of five months at the Iraqi school in New Delhi. It’s a co-education schooling system where 14 learners, who showed interest in this project, were randomly selected. The range of population sample age was 12–15 years (five boys and nine girls, i.e., 35% and 65% respectively). The proficiency level of the language learners was estimated to be pre-intermediate and above as their average formal semester evaluation scores of English subject showed. The medium of instruction at school is Arabic except for English classes it’s English-based most of the time. English for the participants is second as it has to be used outside the classroom where the bigger context is Delhi where English is ‘for most of the population has only ever been a second language’ (Robinson, 2019 ). At the beginning, participants were explained the general framework of the study and given freedom to entirely choose for themselves the content out of the suggested YouTube channel for discussion and writing about for every class. Participants were met three days a week and YouTube was at the heart of the meeting to interact with its content in a watch-take notes-discuss modality using the necessary technological accessories for that matter like an overhead projector, sound amplifier and good internet connectivity for browsing and streaming a given program, in this case the BBC Six-Minute English YouTube channel. It is the program suggested by the researcher as it’s only a six-minute, free of charge weekly show presented in English using the British English. It’s designed and broadcasted by the BBC for intermediate level classes. The rationale behind introducing this channel on YouTube as a tool in ELT classes is that it matches the learners’ overall proficiency level, and these videos use General English—the day-to-day English used exclusively in people’s lives presented in a casual and conversational style that helps learners learn and practice authentic useful English language for everyday situations as in saying, writing or doing something using English.

Pedagogical scenario and task design

At the end of every English class learners were suggested three videos by the instructor (the titles were supposed to be interesting and new to them), and they had to vote (each according to the topic s/he liked) for one of them, and the majority would win for that matter. This new video of BBC six-minute English would be the subject matter for next class. At home, learners start watching to explore and understand the theme on their own. After watching it as many times as needed, learners were asked to write a summary and add their comments about the topic discussed in a given video of this channel. Also, they were encouraged to note down any inquires or doubts about the topic and bring that to class for open discussion. Learners were strongly advised and encouraged to refer to and use the closed captions and its adjustable settings like font size and colour in a given video to check for themselves on the right meaning of the topic/ some sentence or to check (problematic/ unfamiliar) grammatical forms/ vocabulary, word choice, spelling, punctuations, abbreviations, or capitalization. Learners were to do this every time they were given a writing task.

At the start of the writing lesson, the same video with the subtitle turned on was watched by the whole class once again. Then, learners were asked to read out their written summaries and discuss with the class the general idea of the introduced topic. Also, the teacher would basically use the subtitle and its adjustable related settings of font size and colour for their signalling effect (Mayer, 2009 ), which will be projected on the board, to refer to and discuss a particular language point/ inquiry. These language points/ inquiries were mainly those raised directly by the learners themselves, or indirectly when errors were noticed in their writing or even the teacher can pinpoint some language points whenever found noteworthy, relevant and enriching for the learners’ writing. Moreover, learners were encouraged to refer back to the subtitle as a reference point at any time they need to check for themselves and with each other on the right meaning of the topic/ some sentence and check the accuracy of word choice, morphological/ grammatical structures, abbreviations, spelling, punctuations or capitalization.

Data collection

In order to elicit the data from the current population sample, this research adopts a strategy of triangulation through two methods.

Two IELTS-based communicative writing tasks were conducted for the baseline and other two different IELTS writing tasks were conducted as an end line. The tests were given to learners before and after the introduction and integration of the suggested YouTube Channel, BBC Six-Minute English YouTube Channel so as to record their actual writing fluency progress, taking the IELTS standards for the test administration into consideration. Learners are, thus, examined on two communicative tasks each time rather than a single one and the total of two is given as a single line of reference each time for evaluation. This is considered to be more representative of the learners overall communicative competence. The findings of these tests provided the data required to help answering the main research question of this study.

Nonetheless, there is a wide range of likely factors and resources (also known as confounding variables) of English language learning that may co-exist in the learning environment of this research population and hence may impact to varying degrees the improvement of this population’s targeted variable of fluency of writing. Also, this work can only focus on one of these many resources and is mainly interested in the role and impact of ICT multi-media like YouTube as a language learning tool. Therefore, to be able to overcome this challenge and make a valid judgment about other variables which might be playing a role along with the possible YouTube role in the possible development of learners’ writing fluency for this sample group, a quantitative and qualitative online questionnaire was developed according to the specific aims and context of this study, measuring specific potential resources of English language learning in terms of the learners’ daily time range of active exposure to and learning engagement rate with them. The percentages drawn from the population sample responses through the online questionnaire were employed to identify several factors/ independent variables in terms of exposure time and engagement rates with the potential learning sources of English language learning; that would be correlated with the actual linguistic progress learners may make. These percentages are namely the likely potential media or contexts of language learning favoured by learners (e.g. reading and/or listening), preferred mode of exposure (online, offline), modality (text and/or speech), the language skill(s) involved, the type of input material exposed to (as in songs, films, video games, (audio) books). This self-report questionnaire was administered at the end of the five months. The learners’ responses from the survey items were used in the quantitative and qualitative analyses to answer the main research question mentioned above. This questionnaire which was a combination of closed-ended and Likert questions (35 items in total) covered three main areas. First, the participants’ English writing learning experience of both the online and offline English learning resource(s) including YouTube with regards to their engagement rate with and daily time range of exposure to each of these resources (Qs. 4–29 see Appendix A for the questionnaire items related to this area). for the questionnaire items related to this area). Second, the participants’ English writing learning experience with YouTube in particular as an ICT tool for creating the intended smart learning environment with respect to its features of personalization, networking and interactivity, inclusiveness and as a resource for rich language input and engaging multi-media materials for learning and improving your English writing (Qs. 30–33 see Table  8 for the questionnaire items related to this area). Third, learners’ personal and contextual perspectives on the affordances of YouTube videos and how they thought YouTube videos made learning and improving writing easy and interesting according to their experience (Qs. 34 & 35 see Tables  9 & 10 for the questionnaire items related to this area).

Data analysis

The data analysis included both quantitative and qualitative methods. The quantitative methods used in this study were Matched-pairs t-test, Pearson’s correlation, Simple linear regression and frequency distributions. A matched-pairs t-test was used as it is appropriate for a repeated measure design where the same subjects are evaluated under two different conditions such as the case in this study. Pearson’s Correlation was used to explore the relationship between the learners’ writing fluency progress (as the dependent variables for correlation; the data for these variables were elicited from the T. test findings after the integration of YouTube) and the participants’ English writing learning day-to-day experience with some potential online/ offline mono-medium and multimedia learning materials/ environments, including YouTube media with regards to the participants’ engagement rate with and daily time range of exposure to each of these resources. (as the independent variables for correlation; the data for these variables were elicited from the respective questionnaire items Qs. 4–29). Simple linear Regression, focusing primarily on the Likert items which showed a linear relationship, was conducted to identify the variation of the engagement rate with and time range of exposure to YouTube as predictor variables on the writing fluency metrics. Frequency Distributions were used to determine the percentages of learners’ active use of ICT with regards to its features of personalization, networking and interactivity, inclusiveness and as a resource for rich language input and engaging multi-media materials for learning and improving English writing (the data for frequency distribution were elicited from the questionnaire items Qs. 30–33).

In terms of the qualitative method, a simple content analysis consisting of two closed-ended questions was performed. Learners were asked to choose how they thought YouTube videos made learning and improving writing easy and interesting according to their experience. The answers list (for these two questions) learners were to choose from are widely mentioned in the literature of learning writing (except for those relating to closed captions which were investigated in this study) and in line with the research questions of this study (the data for content analysis were elicited from the questionnaire items Qs. 34–35).

Coding the data

As for the test findings, all the participants’ responses ( n. 14) were analysed in terms of writing accuracy objectively using the taxonomy of errors by Dulay, H. et al. ( 1982 ) cited in Ellis and Barkhuizen ( 2005 ) and then coded based on the work by Oshima and Hogue ( 1997 ) and developed in the work of Ferdouse ( 2013 ). Each writing task was rated twice and the inter-rater reliability of the ratings for the 56 writing tasks used in this study was 0.87. Inter-rater reliability rates between 0.75 and 0.9 are good, Koo and Li ( 2016 ). Also, this research used seven metrics or dimensions of various foci of writing fluency of quantitative and qualitative nature widely accepted in the literature as indicators of fluency development (see Ellis, R., 1990 ; Lu, X., 2010 , 2011 ; Lu, X., & Ai, H, 2015 ; Wolfe-Quintero, K., et al., 1998 ; Vaezi & Kafshgar, 2012 ; Fellner and Apple, 2009 ; Van Gelderen, A., & Oostdam, R., 2005 ); these were used as the dependent variables of this study. These dependent variables are of quantity and quality nature. This study used only one quantitative dimension of writing fluency, namely the writing rate which is measured by the number of syllables written per minute in a text. The qualitative dimensions of writing fluency were the number of error-free T. units per text as a sub-dimension of accuracy dimension; lexical diversity (measured by different number of words/total number of words× 100) and lexical density (measured by the number of content words/total number of words× 100) as sub-dimensions of lexical complexity dimension; mean length of T. unit and number of clauses per T.unit as sub-dimensions of syntactic complexity dimension, and organization of ideas which is measured by the overall coherence & cohesion of ideas and task achievement. It should be mentioned that the author used online automatic softwares to conduct the required writing fluency analyses for this study (see Web-based L2 Syntactic Complexity Analyzer by Ai, H., ( n.d. ), Ai, H., & Lu, X. 2013 ; Analyze My Writing ( n.d. ); TAASSC, see Kyle, K. 2016 ; Text Inspector, ( n.d. )). These softwares are cited in the reference list with the respective URLs.

As for the questionnaire findings, all the participants’ responses ( n. 14) were downloaded to a Microsoft Excel sheet and subsequently exported to SPSS version 21.0. With respect to the engagement rate, it was measured on a five-point Likert scale where 1-point was defined as very low engagement and 5-points as very high engagement. As for the exposure time range, it was defined by the number of hours spent in a day using each of these resources for learning English and measured on a five-point Likert scale. Time range of exposure was coded as 0, one hour as 1, two hours as 2, three hours as 3 and more than 3 h as 4.

Quantitative results

Matched-pairs t-test.

In order to be able to check and account for the difference with regards to the baseline compared to the end line datasets of learners’ writing performance objectively, the statistical T. test was used. The means differences between these datasets showed to various degrees some improvement across all the dependent variables set in this work except for the lexical density variable as shown by the means differences in the fluency gain level of writing post the integration of YouTube as an ICT tool for learning and improving writing (Table  1 ) (Details on each variable will be discussed below). Most statistical analyses would use an alpha of 0.05 as the cut-off for the level of significance. If it is found that the p value < 0.05, then the null hypothesis that there is no difference between the means before and after the study can be rejected. The following null hypothesis (H o ) set in this work and its alternative hypothesis (H a ) are as follows:

H o  = exposure to and engagement with ICT educational multimedia like YouTube has no effect on the development of learners’ fluency of language use and expression in writing.

H a  = exposure to and engagement with ICT educational multimedia like YouTube has an effect on the development of learners’ fluency of language use and expression in writing.

The following are the T. test results (see Table 1 ) for the quantitative and qualitative writing fluency dimensions.

Quantitative dimension

Writing rate.

The results from the pre-test ( M  = 5.72, SD  = 2.09) and post-test ( M  = 6.19, SD  = 1.66) writing tasks showed a slight improvement in the learners’ writing fluency in terms of writing rate after the exposure to the suggested ICT tool, t (14) = 1.353, p  = .199.

Qualitative dimension

Number of error-free t. units per text.

The results from the pre-test (M = 1.57, SD = 2.24) and post-test (M = 7.21, SD = 6.14) writing tasks showed good improvement in writing fluency in terms of the number of error-free T.units per text after the exposure to the suggested ICT tool, t (14) = 4.623, p  < .001.

Lexical diversity subdimension of Lexical complexity

The results from the pre-test ( M  = 52.54, SD  = 19.96) and post-test ( M  = 54.86, SD  = 15.11) writing tasks showed a slight improvement in writing fluency in terms of lexical diversity after the exposure to the suggested ICT tool, t (14) = 0.526, p  = .608.

Lexical density subdimension of Lexical complexity

The results from the pre-test ( M  = 47.25, SD  = 4.59) and post-test ( M  = 44.41, SD  = 2.75) writing tasks showed no improvement in writing fluency in terms of lexical density after the exposure to the suggested ICT tool, t (14) = 2.816, p =  .015.

Mean length of T. unit subdimension of Syntactic complexity

The results from the pre-test ( M  = 9.97, SD  = 1.95) and post-test ( M  = 10.48, SD  = 2.22) writing tasks showed a slight improvement in writing fluency in terms of syntactic complexity after the exposure to the suggested ICT tool, t (14) = 1.063, p  = .307.

Number of clauses per T. unit subdimension of Syntactic complexity

The results from the pre-test ( M  = 1.26, SD  = 0.32) and post-test ( M  = 1.34, SD  = 0.28) writing tasks showed a slight improvement in writing fluency in terms of syntactic complexity after the exposure to the suggested ICT tool, t (14) = 0.960, p  = .354.

Organization of ideas

The results from the pre-test ( M  = 4.07, SD  = 1.07) and post-test ( M  = 6.36, SD  = 1.28) writing tasks showed a slight improvement in writing fluency in terms of the number of error-free T.units per text, t (14) = 8.000, p  < .001.

To summarize the T. test results, all dependent variables (except for the lexical density variable) set as indicators of fluency of writing in this study have to various degrees shown some improvement after five months of focused exposure to YouTube as indicated by their means differences in the T. test results. However, only two of them were statistically significant, namely, the number of error-free T. units per text and organization of ideas.

Correlation coefficient results (the critical value approach)

Pearson’s Correlation was used to explore the strength and the direction of the relationship between the learners’ writing fluency which was based on the T. test findings after the integration of YouTube and the participants’ English writing learning day-to-day experience with some potential online/ offline mono-medium and multimedia learning materials, including YouTube media with regards to the participants’ engagement rate with and daily time range of exposure to each of these resources. With regards to the strength of association between variables, the critical value for Pearson’s |r| was set at the value of 0.05 = 0.532 (as a level of significance for a two-tailed test given by Pearson table of critical values) with a degree of freedom df = N-2, df = 14–2 = 12. So, any correlation coefficient value falling below the 0.532 was considered insignificant and any correlation coefficient value above the 0.532 was seen as significant (only the significant correlation coefficient values were highlighted in the tables below), where the higher the r value or closer to | + 1/− 1| the stronger the correlation is between any two variables.

As for the direction (i.e., negative = −ve / positive = +ve) of the relationship between variables, all seven dependent variables in each of the correlation tables below should be positively correlated with the independent variables as these particular dependent variables were dealing with factors where higher rates or values reflect higher or better outcome on the part of the learners so that the higher the value of each correlation coefficient means the better the performance or the more fluent the writing is. No negative correlation between any of these variables should be predicted for this work.

The critical value for Pearson’s |r| was set at the level of 0.05 = 0.532.

The results of the correlation analysis were broadly divided into two main categories — online Vs. offline learning.

Online learning

The correlation coefficient results of the learners’ engagement rate with the respective online mono-medium and multimedia learning sources showed various values of which the majority were insignificant in terms of correlation, especially in the mono-medium learning environment. Nevertheless, the significant correlation values in terms of both strength and direction were those in the multimedia environment of YouTube, video games and audio books. (see Table  2 ). These significant correlations were as follows:

One strong +ve correlation (r = .763) for the multimedia provided by video games;

One strong +ve correlation (r = .606) for the multimedia provided by films;

One moderate +ve correlation (r = .575) for the multimedia provided by audio books; and

One moderate +ve correlation (r = .598), two strong +ve correlations (r = .606, .732) and two very strong +ve correlations (r = .867, .844) for the multimedia provided by the YouTube.

These results showed that the correlations of learners’ engagement are significant on the multimedia side rather than the mono-medium one. However, more specifically there are more significant and stronger correlation values with the multimedia by YouTube against most of the set metrics of writing fluency than the correlation values with multimedia by video games, films or audio books.

This may indicate that learners were engaging with the multi-media materials by YouTube more and to a greater extent than the rest of other learning sources in their online learning environment and that the multi-media materials are preferred over the mono-medium materials.

The correlation coefficient results of the learners’ time range of exposure to the respective online mono-medium and multimedia learning sources showed that there were insignificant correlations from the majority of the resources except for the YouTube and video games results (see Table 3 ) which were as follows:

One strong +ve correlation (r = .699) for the multimedia provided by video games;

Four strong +ve correlations (r = .783, .734, .767, .761) for the multimedia provided by YouTube.

The results of time range of exposure (Table 3 ) seem to conform with the above results of engagement rate (Table 2 ) that while learners were giving far greater amounts of their learning time and attention in the multimedia leaning environment, they gave little or no time and attention in the mono-medium learning environment. This may suggest that learners were more inclined towards the multi-media learning environment than the mono-medium environment for learning and improving writing online; more specifically, when compared to other online multi-media learning sources, YouTube as multimedia learning tool cum environment was preferred over other learning multimedia in as far as learning and improving writing is concerned.

Offline learning

The correlation coefficient results of the learners’ engagement rate with the respective offline mono-medium and multimedia learning sources showed insignificant correlation values across all the variables except for one strong +ve correlation value for video games (r = .763). (see Table  4 ).

These insignificant results may suggest that learners were not engaging their learning in the offline mono-medium and multimedia learning environments at all.

The correlation coefficient results of the learners’ time range of exposure with the respective offline mono-medium and multimedia learning sources showed that there are insignificant correlations from most of the resources in terms of time of exposure except for the reading of books variable results which are as follows: (see Table 5 )

Two strong +ve correlations (r = .713, .774) and one very strong +ve correlation (r = .842).

This may suggest that learners were giving a good deal of their learning time to reading books in the offline mono-medium leaning environment.

To summarize the correlation results of the above mentioned potential online/ offline mono-medium and multimedia learning materials/ environments, including YouTube media with regards to the participants’ engagement rate with and daily time range of exposure to each of these resources, it can be concluded with statistical evidence that there is a strong positive correlation between the writing fluency performance shown by this group of learners and multi-media (rather than mono-medium) learning environments found online (rather than offline) such as the suggested ICT YouTube tool, where text can be optionally available along with speech. Moreover, multi-media materials/ environments like YouTube as an example are more engaging compared to other language learning sources and hence learners were spending much more time using them which could have led to a more effective learning in their writing fluency; this may suggest the usefulness of ICT tools like YouTube as an example which have resulted in a more fluent use of written language in some (but not all) dimensions of writing fluency over time as learners could with the help of ICT handle their learning efficiently. However, from a statistical perspective, it is important to stress that correlation between any two variables does not mean causation but only that there is a relationship between them such as the kind of positive linear relationship which exists between engagement rate and time range of exposure in relation to the writing fluency metrics of this study (see Fig.  1 . a & b). These correlational findings which revealed such positive linkages among the data set encouraged the employment of the following regression analysis.

a Sample of strongly positive linear correlation. b Sample of moderately positive linear correlation

Regression analysis results

Regression analysis estimated the variation produced by engagement rate with and time range of exposure to YouTube as a source of learning and improving writing on the writing fluency metrics in this study.

The regression analysis results for the engagement rate showed a range of low, moderate and high R Square values (only the moderate and high values were highlighted in the table below). The moderate R Squared values were ( R 2  = .367, .358, .536) and the high values were ( R 2  = .751, .712). R squared values represented the proportion of the variance for the respective writing fluency metrics explained by the engagement rate as a predictor. (see Table  6 ).

Also, most, but not all, of the actual coefficients p values (used to test the null hypothesis where the coefficient is equal to zero i.e., meaning there is no effect while a low p value < 0.05 indicating that the null hypothesis can be rejected) shown in this table were too small, p  < 0.05. These values also signified a significant linear relationship between the engagement rate with YouTube against their respective writing fluency metrics under study (see Fig. 1 . (a)). The Significance F. values (these values expressed the results of the F. statistic used to measure the significance of the model and the level of significance) were significant as they were well below the P  < 0.05.

So, it may be suggested with statistically significant P and R Squared values of this regression that the engagement rate with the suggested YouTube channel can variably (moderate-high range) explain between 0.35 to 0.75% of the fluency improvement in the learners’ writing for the corresponding writing fluency variables.

The regression analysis results for the time range of exposure showed a range of low and moderate R Square values (only the moderate values were highlighted in the table below). R squared values represented the proportion of the variance for the respective writing fluency metrics explained by the exposure time rate as a predictor (see Table 7 ). Also, most, but not all, of the actual coefficients p values shown in this table were too small, p  < 0.05. These values signified a significant linear relationship between the time range of exposure to YouTube against their respective writing fluency metrics under study (see Fig. 1 . (b)). The Significance F . values were significant as they were well below the P  < = .05.

So, it may well be suggested that with statistically significant P and R Squared values of this regression that the time range of exposure to the suggested YouTube channel can variably (low-moderate range) explain between 0.53 to 0.61% of the fluency improvement in the learners’ writing.

Frequency distributions

Frequency Distributions were run on data responses of the questionnaire for the following dichotomous (Yes/No) questions (Table 8 ) to determine the percentages of learners’ active use of ICT with regards to its features of personalization, networking and interactivity, inclusiveness and as a resource for rich language input and engaging multi-media materials for learning and improving English writing. Percentages of use for these ICT features can also be found in (Fig.  2 ).

Features of ICT technology like YouTube Vs. Percentage of active use by this group of learners while learning English

From the frequency distributions analysis, it can be seen that all of the participants used the YouTube closed captions and its adjustable related settings of font size, colour, opacity and playback speed when learning and improving writing (Question 30 above). (Question 31 above) showed that nearly all the participants shared with other learners interesting YouTube video materials for learning writing and learnt from what others or YouTube itself suggested them for that matter. (Question 32 above) showed that more than half of the participants used what they thought and learnt through these YouTube videos whenever they wrote in English. The majority of the participants did subscribe and passionately follow the YouTube channels due to their rich language input and engaging multi-media materials for learning and improving your English writing (Question 33 above). Taken together, analyzing these frequency distributions indicated that the participants actively used ICT tools like YouTube with regards to its features of personalization, networking and interactivity, inclusiveness and as a resource for rich language input and engaging multi-media materials for learning and improving English writing (see Table 8 ).

Qualitative results

Content analysis.

In general, the results of this qualitative analysis supported the quantitative findings and brought more information about learners’ personal and contextual perspectives on the affordances of YouTube videos and how they thought YouTube videos made learning and improving writing easy and interesting according to their experience. (see Tables 9 and 10 ). Participants were asked to choose from the answers given to these closed-ended questions what they thought applied to them according to their personal learning experience. The following tables are learners’ responses (given in numbers and percentages) in the affirmative for the two-closed questions (Q.34 & 35) in the questionnaire.

Percentages of participants’ responses in this simple content analysis varied over how they thought the affordances of YouTube videos made learning and improving writing easy and interesting. Nevertheless, the majority of participants, in response to each of the above statements (see Tables 9 and 10 ), thought that the videos aided in different ways in the development and learning of writing. Overall, the participants found that the affordances of YouTube videos made learning and improving writing easy and interesting with respect to the above listed affordances of YouTube (see Tables 9 and 10 ).

Discussions

This study examined the potential role and impact smart of learning environment of ICT tools like YouTube on learners’ fluency of language use and expression in their daily written communication. Three main areas related to the main research questions were analysed in this study.

the participants’ English writing learning experience of both the online and offline English learning resources, including YouTube with regards to their engagement rate with and daily time range of exposure to each of these resources;

the participants’ actual use of YouTube in particular as an ICT tool for creating the intended smart learning environment with respect to its features of personalization, networking and interactivity, inclusiveness and as a resource for rich language input and engaging multi-media materials for learning and improving English writing; and

participants’ personal and contextual perspectives on the affordances of YouTube videos and how they thought YouTube videos made learning and improving writing easy and interesting.

Previous research in this area seemingly devoted considerable effort and emphasis on the impact of YouTube usage in the classroom. In this regard, this research has gone a step further by examining the potential impact of such usage on learners’ writing fluency in particular. The quantitative findings of the T. Test (Table 1 ) clearly show some progress in the writing fluency post the integration of YouTube as a tool of language learning over the course of five months for this group of learners. Nonetheless, the T. Test results also show that there is a statistically significant difference only in terms of the number of error-free T. units and organization of ideas but not across all the outcome variables which were used as indicators of writing fluency in this work. The findings of this T. Test support previous studies such as those by Pratiwi ( 2011 ) and Anggraeni ( 2012 ) who reported that YouTube videos help the students to explore main ideas, organize ideas, choose right words to create sentences and paragraphs, produce grammatically correct sentences and use mechanic (punctuation and spelling) in writing. Thus, YouTube is effective in helping the students to better write, quantity and quality-wise, in English.

Furthermore, the results show a clear tendency to both engage and spend more time with the smart learning environment of the online multimedia materials (text and speech) like YouTube to a higher degree and comparatively to a lesser degree with (audio) books, video games, films rather than with online mono-medium materials (text/speech) as it can be seen in the above-Tables 2 and 3 ). However, there is evidence about learners’ offline mono-medium time spent on reading books, including their school syllabus books and assigned readings but with little or no corresponding engagement rate; conversely, learners showed some engagement in offline mono-medium video games but with little or no exposure time (Tables  4 and 5 ). In other words, multi-media rather than mono-medium materials in the online environment rather than offline, as it is the case in this study, brought by ICT technology like YouTube seem to be all the more engaging (the highest among the group of learning resources) for learners so much so that their learning time using this tool seems to be the highest vis-à-vis time ranges given to other learning resources. In this respect, the results of correlation coefficient (Tables  2 and 3 ) and regression analyses (Tables  6 and 7 ) bring in a strong evidence to suggest that the difference in writing fluency performance of this group post the integration of YouTube can clearly be explained by the existing positive linear relationship (Fig. 1 a, b) between, on the one hand, the learners’ engagement rate with and exposure time to the smart learning environment of YouTube as a source of multimedia language learning input and their fluency improvement in writing, on the other. Such positive linear relationship can be seen more clearly in the case of YouTube media and writing fluency metrics than with any other potential learning resources and writing fluency metrics in this study as indicated by the correlation results (see Tables 2 , 3 , 4 and 5 ). These findings are in line with previous studies which found that YouTube usage was linked to students being more engaged, experiencing improved critical thinking ability and greater depth of learning (Clifton & Mann, 2011 ).

ICT multimedia learning materials can do a good job for the language learners at different stages of the learning experience as it can not only bring to them an increasing number of authentic language learning materials but also such learning materials can enhance their learning with the ICT capacity of personalization, networking, inclusiveness and engaging and authentic language input. The percentages of ICT features usage by learners (Fig. 2 .) suggest that multimedia learning is successfully driving learners to help themselves down the process of language learning through availing the various built-in ICT features and its supplied multi-mediated language input. In this regard, the quantitative findings seen in the frequency distributions (see Table 8 ) reveal the participants’ actual use of YouTube for creating the intended smart learning environment with respect to its features of personalization, networking and interactivity, inclusiveness and as a resource for rich language input and engaging multi-media materials for learning and improving English writing. This can reflect the amount of engagement, interaction and enhancement of direct and indirect learning outcomes. In a similar vein, Mayora ( 2009 ) focused on the use of YouTube for eliciting improved writing by language learners and also explored how authenticity, interaction, and motivation are inter-twined. He concluded that certain features of YouTube, including the written comments and the possibility for students to express their ideas by constructing meaning through the stimulus of the videos can improve student’s writing skills through authentic interaction.

The qualitative responses seen in Tables 9 and 10 demonstrate the positive impact of the affordances of YouTube videos which can aid in different ways in the development and learning of writing, i.e., the affordances of YouTube videos made the process of learning and improving writing easy and interesting. Such affordances of YouTube reflected what has been recently reported in language education (Izquierdo et al., 2015 ) that ICT makes it easier for learners to access language materials, stressing an existing correlation between second language learning and the use of multi-media materials in a computer-enhanced language learning milieu showing an impact on learning behaviour with increased motivation.

Taken together, the quantitative and qualitative results of this work confirm previous research and indicate that multimedia brought by ICT technology as a source of language learning can be reliably effective for the optimization of language skills including writing as long as some conditions in the learning environment are met, mainly comprehensibility of input, sufficient exposure time devoted by the learner to the learning material and high engagement level with the material rather than the technology itself. (see ; Alvarez-Marinelli et al., 2016 ; Hung, Huang, & Hwang, 2014 ; Kelsen, 2009 ; Malhiwsky, 2010 ; Warschauer & Grimes, 2007 ).

While learning on YouTube, multimedia learning was actually activated in class and learners were most of the class time observed to be listening attentively to the content of the videos and more importantly fully attracted to the subtitle on the screen which was adjusted size- and colure-wise for explicit signaling effect purposes. Learners did this whenever they had/ were asked to check the right meaning of a particular sentence; they had some doubt/ were asked about an answer to a question in the video or had some inquiry related to language point mentioned in some video like unfamiliar or problematic word meaning/ spelling or grammatical structure). In this regard, based on the basic premise of multimedia learning on the advantage of presentations including signaled text, learners can actively integrate new knowledge into a coherent linguistic and mental model. In other words, linguistic elements need to be linked to some visual stimuli so as to assist learners’ storage of the new linguistics elements in their long-term memory; this may indicate how YouTube media help learners’ errors (at least to the extent this study findings could demonstrate) move from error to correct. These observations of what the learners’ experienced in this study regarding the captioning support and add value to previous studies (Winke et al., 2010 ) which also mentioned “a number of observations about the use of captions that captions are beneficial because they result in greater depth of processing by focusing attention, reinforce the acquisition of vocabulary through multiple modalities, and allow learners to determine meaning through the unpacking of language chunks”. In this respect, learners writing has been improved due to the clarity and comprehensibility of the language input in these YouTube videos which could have been assisted through the affordances of ICT multimedia, i.e., enriching for the learners’ language as in helping learners write fluently (quantity- as well as quality-wise). ( Appendix B ) provides two writing samples for two learners showing the kind of improvement in the quality and quantity of writing pre and post the integration of YouTube media for learning and improving writing, like the organisation of ideas and number of errors which have improved besides some new constructions or language chunks that learners seem to have picked from the videos probably by way of chunk learning (it should be mentioned that these chunks were used frequently in these videos). Some examples of these new constructions, which were observed in the learners’ writing post the integration of YouTube media, are as well as, I appreciate it, twice a day/ week, a day, there’s nothing to worry about, as well, even if, again and again, but that doesn’t mean, but anyway. What is more, the quantitative findings (Table 1 ), which have shown the actual participants’ writing development, give us some fair idea how some of the learners’ disfluencies have become fluent due to the improvement of the above-mentioned writing metrics dealt with in this study.

Even though cognitive advantages of multimedia learning have been widely recognized and adopted by researchers and practitioners across a wide range of disciplines, there is a concurred view among researchers that multimedia may also impede learning and increase cognitive load if not appropriately designed (see Kozma, 1994 ; Lee, Plass, & Homer, 2006 ; Mayer & Moreno, 2003 , as cited in Cook et al., 2008 ). In this respect, while the captioning might be largely benefiting learners’ in certain contexts (see Yoshino et al., 2000 ; Winke et al., 2010 ), it could be creating cognitive overload on the learners’ working memories at other times as the combination of both speech and text (as it is the case in this experiment) may overwhelm learners’ visual channels according to the dual-channel and limited-capacity assumptions of Mayer’s cognitive theory of multimedia learning (Mayer, 2009 ). This study, hence, suggests a trade-off learning/ teaching strategy, i.e., including the captioning when and as much as required so that it doesn’t overtax the learners’ working memory resources. “Taking this into account, different signalling techniques should be used to gain the learners’ attention and reduce extraneous processing when aiming at focusing on specific linguistic aspects in multimedia presentations incorporated into traditionally implemented classroom practices” (Matus, 2018 ).

Nonetheless, while this study stresses the significance of captioning and its adjustable settings like font size and colour for learning and improving language, many interesting questions related to captioning would immediately arise like if we remove them would we get the same results or otherwise this would change them? Is it the captioning/font/frequency of use/ length of caption/repetition of caption - are all or any of these important for changing the results? Eye tracking would tell us if students are looking at the captions at all? But then what if they look at the same video more than once, (considering Constant Reverse Navigation Pattern of working memory capacity, Graf, Lin, Kinshuk Chen, & Yang, 2009 ) are they still reading the captions? These are particularly relevant and intriguing questions but beyond the scope of this paper and hence deserve the attention and effort in further possible studies.

To conclude with an important statistical note, even though all dependent variables (except for the lexical density variable) set as indicators of fluency in writing have to various degrees shown some improvement in writing fluency as indicated by their means differences in the paired-sample T. test results after five months of focused exposure to YouTube, only two of them consistently showed statistically significant results across the three testing and analysis models (refer to T. test, Pearson correlation and Regression analyses results) used in this work; these are the number of error-free T. units per text, reflecting the accuracy dimension of fluency; and the organisation of ideas of learners, reflecting the coherence and cohesion dimension of fluency. However, the other variables (except for the lexical density variable which did show not any improvement) showed improvement but with no statistical significance in this population sample. Therefore, it can be concluded that there is no sufficient evidence to prove that a significant difference exists between the learners’ performance before and after the exposure to the suggested ICT tool as far as the fluency in writing is concerned; hence, the null hypothesis (H o ) set in this work cannot be entirely rejected nor the alternative hypothesis can be confidently accepted. Such common statistical insignificance and inconsistency on the part of dependent variables may plausibly suggest that more data is needed and further research is required to investigate the issue of writing fluency and to validate the results of this research.

Limitations

One major challenge was learners’ worry about the accuracy over the fluency of their written performance. Therefore, they were reassured that errors are not “signs of inhibition” which need to be eradicated but rather as strategies of learning and as perfectly natural aspects of second language acquisition Corder, S. P. ( 1967 ). However, learners would be reminded of the significance of striking a balance between accuracy and fluency in their formal and informal production as these two aspects of writing are closely related that they are inseparable. Another limitation is related to creating learning environments that are adaptive and responsive to specific learners in specific situations is clearly possible, but as the New Media Consortium’s 2016 Horizon Report for Higher Education indicates personalized learning is a significant challenge (Adams Becker, Freeman, Giesinger Hall, Cummins, & Yuhnke, 2016 ). In this respect, the problem of “one size fits all” when given two or three suggested episodes of the YouTube BBC language show for learners to pick one from to be the crux of the matter for the next class, naturally sometimes few learners would opt for a different topic from what the majority of the class have agreed on to be discussed. In such cases the selected by those few learners but not won by the majority topics would be noted and again re-suggested for other classes to ensure diversified learning materials. Luckily the smart leaning environment could help with such challenges given the multitude of what now became inherent characteristics and advantages of this learning environment and the plethora of increasingly different learning topics and materials for everyone.

To conclude, this research evidently shows that the smart learning environment of ICT multimedia technology as a source for language learning with its multiple handy features can efficiently drive a range of desired effects for the optimization of writing fluency of language use and expression in the language learners’ daily written communication. This research, both informed by the its results and other researches in the literature supports the fact that such effects could be effectively driven as long as the essential factors in the learning environment like high engagement, sufficient exposure time, comprehensibility of learning input (on the part of the language learner) and enhancement and intelligibility of learning input (on the part of the environment/ learning materials) are provided by its multi-mediated input, so much so that successful learning is due at any moment. Also, compared to other sources of language learning in the learners’ environment, multimedia educational tools developed by ICT like the widely known platform YouTube, being preferred over other learning sources, can be more effective and thus strongly recommended equally for language learners and teachers where optimization of writing fluency is the target of learning. Multimedia materials through ICT technology has made the formal and informal experience of learning more effective and interactive as it is more adjustable, shareable and retainable.

Availability of data and materials

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Abbreviations

Information and Communication Technology

Multi-media Language Materials

British Broadcasting Corporation

English Language Teaching

International English Language Testing System

Null Hypothesis

Alternative Hypothesis

Computer Assisted Language Learning

Multi-media Instruction

Social Networks

Adams Becker, S., Freeman, A., Giesinger Hall, C., Cummins, M., & Yuhnke, B. (2016). NMC/CoSN horizon report: 2016 K-12 edition . Austin: The New Media Consortium.

Google Scholar  

Ai, H. (n.d.). Web-based L2 Syntactic Complexity Analyzer. https://aihaiyang.com/software/l2sca/ . Accessed on 2 Jan 2020.

Ai, H., & Lu, X. (2013). A corpus-based comparison of syntactic complexity in NNS and NS university students' writing. In A. Díaz-Negrillo, N. Ballier, & P. Thompson (Eds.), Automatic treatment and analysis of learner corpus data , (pp. 249–264). Amsterdam/Philadelphia: John Benjamins.

Alvarez-Marinelli, Blanco, M., Lara-Alecio, R., Irby, B.J., Tong, F., Stanley, K., & Fan, Y. (2016). Computer assisted English language learning in Costa Rican elementary schools: an experimental study, Computer Assisted Language Learning, 29 (1), 103–126, https://doi.org/10.1080/09588221.2014.903977 .

Analyze My Writing. (n.d.). (Lexical Density Calculator online). http://www.analyzemywriting.com/index.html . Accessed May 2020.

Anggraeni, S.N. (2012). Optimizing the Use of Youtube Video to Improve Students’ Competence in Writing Procedure Text (A Classroom Action Research at the Tenth Grade Students of SMA N Kebakkramat in Academic Year of 2011/2012). Sebelas Maret University. Unpublished Thesis.

Annetta, L. A., Minogue, J., Holmes, S. Y., & Cheng, M. T. (2009). Investigating the impact of video games on high school students’ engagement and learning about genetics. Computers in Education , 53 , 74–85. https://doi.org/10.1016/j.compedu.2008.12.020 .

Article   Google Scholar  

Aoki, K. (2010). The use of ICT and e-learning in higher education in Japan. World Academy of Science, Engineering and Technology international journal of educational and pedagogical sciences , 4 , 6.

Astin, A. (1984). Student involvement: A developmental theory for higher education. Journal of College Student Personnel , 25 (4), 297–308 https://www.middlesex.mass.edu/ace/downloads/astininv . Accessed 1 Jan 2020.

Astin, A. W. (1993). What matters in college? Four critical years revisited . San Francisco: Jossey-Bass https://www.researchgate.net/publication/242362064_What_Matters_in_College_Four_Critical_Years_Revisited . Accessed 20 Jan 2020.

Atasoy, A., & Temizkan, M. (2016). Evaluation of secondary school students’ writing fluency skills. Educational Sciences: Theory & Practice , 16 , 1457–1484 https://files.eric.ed.gov/fulltext/EJ1115153 . Accessed 2 Jan 2020.

Benek-Rivera, J., & Matthews, V. E. (2004). Active learning with jeopardy: Students ask the questions. Journal of Management Education , 28 , 104–118 http://www.upv.es/gie/repositorioIEMA/ek-Rivera&Mathews2004Active%20Learning%20With %20Jeopardy . Accessed 15 Feb 2020.

Benson, P. (2001). Teaching and researching autonomy in language learning . Harlow: Pearson Education http://www4.pucsp.br/inpla/benson_artigo . Accessed 14 Feb 2020.

Bertin, J.-C., Grave, P., & Narcy-Combes, J.-P. (2010). Second-language distance learning and teaching: Theoretical perspectives and didactic ergonomics . USA: IGI Global.

British Council. (2020a). https://www.teachingenglish.org.uk/article/comprehensible-input#sidr-0 ; Accessed 1 Mar 2020

British Council. (2020b). https://www.teachingenglish.org.uk/article/exposure ; Accessed 10 Feb 2020

Brophy, J. (1983). Conceptualizing student motivation. Educational Psychologist Journal , 18 , 200–215.

Buchem, I., & Pérez-Sanagustín, M. (2013). Personal learning environments in smart cities: Current approaches and future scenarios. eLearning Papers, 35 . http://openeducationeuropa.eu/sites/default/files/asset/In-depth_35_1.pdf .

Chen, P. D., Lambert, A. D., & Guidry, K. R. (2010). Engaging online learners: The impact of web-based technology on college student engagement. Computers in Education , 54 , 1222–1232. https://doi.org/10.1016/j.compedu.2009.11.008 .

Clifton, A., & Mann, C. (2011). Can YouTube enhance student nurse learning? NurseEducation Today , 31 (4), 311–313. https://doi.org/10.1016/j.nedt.2010.10.004 .

Connell, J. P. (1990). Context, self, and action: A motivational analysis of self-esteem processes across the life-span. In D. Cicchetti (Ed.), The self in transition: From infancy to childhood , (pp. 61–97). Chicago: University of Chicago Press.

Cook, A., Zheng, R., & Blaz, J. (2008). Measurement of cognitive load during multimedia learning activities. Cognitive Effects of Multimedia Learning. , 34–50. https://doi.org/10.4018/978-1-60566-158-2.ch003 .

Corder, S. P. (1967). The significance of Learner’s errors IRAL 5/1967. 161-170. Accessed 22 Apr 2020.

Dulay, H., M. Burt, and S. Krashen. (1982). Language Two. New York: Oxford University Press.

Ellis, R. (1990). 'Individual styles in classroom second language development' In J. de Jong and D. Stevenson (eds.): Individualizing the Assessment of the Language Abilities . Clevendon, Avon: Multilingual Matters.

Ellis, R. (2002). Second language acquisition . Oxford: Oxford University Press.

Ellis, R., & Barkhuizen, G. (2005). Analysing learner language . New York: Oxford University Press.

Fellner, T., & Apple, M. (2009). Developing writing fluency and lexical complexity with blogs. The Jalt Call Journal, 2 (1), 15–26.

Ferdouse, F. (2013). Learning from mistakes: Using correction code to improve students writing skill in English composition class. Stamford Journal of English , 7 , 62–86. https://doi.org/10.3329/sje.v7i0.14463 Accessed 1 Jan 2020.

Finn, J. D. (1989). Withdrawing from school. Review of Educational Research , 59 , 117–142.

Global eSchools and Community Initiative (GeSCI). (2007). Concept note: Developing a model for inclusive education and assistive technology appropriate for teaching and learning contexts in developing countries. Available at http://www.gesci.org/old/files/docman/model_ie_at.pdf .

Graf, S., Lin, T., Kinshuk Chen, N. S., & Yang, S. J. H. (2009). Learning styles and cognitive traits—Their relationship and its benefits in web-based educational systems. Computers in Human Behavior , 25 (6), 1280–1289.

Grzeszczyk, K. B. (2016). Using multimedia in the English language classroom Kamila Barbara http://www.worldscientificnews.com/wp-content/uploads/2015/10/WSN-433-2016-104-1571. Accessed 20 Apr 2020 .

Gurbin, T. (2015). Metacognition and technology adoption: Exploring influences. Procedia - Social and Behavioral Sciences , 191 , 1576–1582.

Hung, C., Huang, I., & Hwang, G. (2014). Effects of digital game-based learning on students’ self-efficacy, motivation, anxiety, and achievements in learning mathematics. Journal of Computer Education , 1 , 151–166. https://doi.org/10.1007/s40692-014-0008-8 .

Izquierdo, J., Simard, D., & Garza, M. G. (2015). Multimedia instruction & language learning attitudes: A study with university students. Revista Electrónica de Investigación Educativa , 17 (2), 101–115 http://redie.uabc.mx/vol17no2/contents-izqsimard.html. Accessed 30 Mar 2020 .

Junco, R. (2012). The relationship between frequency of Facebook use, participation in Facebook activities, and student engagement. Computers & Education, 58 , 162–171. https://doi.org/10.1016/j.compedu.2011.08.004 .

Kelsen, B. (2009). Teaching EFL to the iGeneration: A survey of using YouTube as supplementary material with college EFL students in Taiwan. CALL-EJ. , 10 .

Koo, T. K., & Li, M. Y. (2016). A guideline of selecting and reporting Intraclass correlation coefficients for reliability research. Journal of Chiropractic Medicine , 15 (2), 155–163 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4913118/ Accessed 3 Apr 2020.

Kozma, R. B. (1994). Will media influence learn-ing? Reframing the debate. Educational Technol-ogy Research & Development, 42 , 7–19.

Kramsch, C. (1999). Teaching text and context through multimedia. Language, Learning and Technology , 2 (2), 31–41 https://scholarspace.manoa.hawaii.edu/bitstream/10125/25041/1/02_02_article1 . Accessed 22 Apr 2020.

Krashen, S. D. (1985). The input hypothesis: Issues and implications . New York: Longman.

Krashen, S. D., Long, M. A., & Scarcella, R. C. (1979). Age, rate, and eventual attainment in second language acquisition. TESOL Quarterly , 13 , 573–582 https://www.academia.edu/40603904/Age_Rate_and_Eventual_Attainment_in_Second_Language_Acquisition . Accessed 8 Mar 2020.

Kunschak, C. (2004). Language Variation in Foreign Language Teaching: On the Road to Communicative Competence . Frankfurt/Main: Peter Lang Publishing.

Kyle, K. (2016). Measuring syntactic development in L2 writing: Fine grained indices of syntactic complexity and usage-based indices of syntactic sophistication (Doctoral Dissertation). Retrieved from  http://scholarworks.gsu.edu/alesl_diss/35 .

Lee. H., Plass, J. L., & Homer, B. D. (2006). Optimizing cognitive load for learning from computer-based science simulations. Journal of Educational Psychology, 98 (4), 902–913. (2) (PDF) Measurement of cognitive load during multimedia learning activities. Available from: https://www.researchgate.net/publication/288581585_Measurement_of_cognitive_load_during_multimedia_learning_activities .

Libbrecht, P., Müller, W., & Rebholz, S. (2015). Smart learner support through semi-automatic feedback . https://doi.org/10.1007/978-3-662-44447-4_8 .

Book   Google Scholar  

Long, M. H. (1983). Native speaker/ non-native speaker conversation and the negotiation of comprehensible input. Applied Linguistics , 4 , 126–141 https://core.ac.uk/download/pdf/32302617 . Accessed 24 Jan 2020.

Long, M. H. (1985). Input and second language acquisition theory. In S. M. Gass, & C. G. Madden (Eds.), Input in second language acquisition , (pp. 377–393). Rowley: Newbury House.

Lu, X. (2010). Automatic analysis of syntactic complexity in second language writing. International Journal of Corpus Linguistics , 15 (4), 474–496.

Lu, X. (2011). A corpus-based evaluation of syntactic complexity measures as indices of college-level ESL writers's language development. TESOL Quarterly , 45 (1), 36–62.

Lu, X., & Ai, H. (2015). Syntactic complexity in college-level English writing: Differences among writers with diverse L1 backgrounds. Journal of Second Language Writing , 29 , 16–27.

Malhiwsky, D. (2010). Student achievement using web 2.0 technologies: A mixed methods study .

Matus, F. (2018). Effects of the Signalling principle on EFL learning: A study of explicit presentation of frequent grammar mistakes using an adapted functional teaching approach. Athens Journal of Philology , 5 , 97–120. https://doi.org/10.30958/ajp.5-2-2 .

Mayer, R. (2005). Multimedia learning: GuidingVisuospatial thinking with instructional animation. In P. Shah, & A. Miyake (Eds.), The Cambridge handbook of visuospatial thinking , (pp. 477–508). Cambridge: Cambridge University Press.

Mayer, R. E. (2002). Multimedia learning https://www.sciencedirect.com/science/article/pii/S0079742102800056. Accessed 23 Feb 2020 .

Mayer, R. E. (2009). Multimedia learning . New York: Cambridge University Press.

Mayer, R. E., & Moreno, R. (2003). Nine ways to reduce cognitive load in multimedia learning. Educational Psychologist, 38 (1), 43–52.

Mayora, C. A. (2009). Using YouTube to encourage authentic writing in EFL classrooms. TESL Reporter , 42 (1), 1–12.

McKay, S. L. (2002). Teaching English as an International Language: Rethinking Goals and Approaches. Oxford: Oxford University Press.

Mikulecký, P. (2012). Smart environments for smart learning. In Proceedings of DIVAI 2012–9th International Scientific Conference on Distance Learning in Applied Informatics (pp. 213–222).

Mitterer, H., & McQueen, J. M. (2009). Foreign subtitles help but native-language subtitles harm foreign speech perception. PLoS One , 4 (11), e7785. https://doi.org/10.1371/journal.pone.0007785 Accessed 2 Feb 2020.

Mukhaini, E., Al-Qayoudhi, W., & Al-Badi, A. (2014). The adoption of social networking in education: A study of the use of social networks by higher education students in Oman. Journal of International Education Research https://clutejournals.com/index.php/JIER/article/view/8516/8523. Accessed 6 Mar 2020 .

Natriello, G. (1984). Problems in the evaluation of students and student disengagement from secondary schools. Journal of Research and Development in Education , 17 , 14–24.

Oshima, A., & Hogue, A. (1997). Introduction to academic writing . Addison Wesley: Longman.

Pannell, J., Partsch, F., & Fullver, N. (2017). The output hypothesis: From theory to practice. TESOL working paper series, 15 , (pp. 126–159) https://www.hpu.edu/research-publications/tesol-working-papers/2017/2017-new-with-metadata/06pannellpartschfuller_output . Accessed 10 Apr 2020.

Patera, M., Draper, S., & Naef, M. (2008). Exploring magic cottage: A virtual reality environment for stimulating children’s imaginative writing. Interactive Learning Environments , 16 , 245–263. https://doi.org/10.1080/10494820802114093 Accessed 4 Feb 2020.

Peregoy, S. F., & Boyle, O. F. (2005). Reading, writing, and learning in ESL: A resource book for K12 teachers Pearson/Ally & Bacon.

Plass, J. L., & Jones, L. C. (2005). Multimedia learning in second language acquisition. In R. E. Mayer (Ed.), The Cambridge handbook of multimedia learning (pp. 467–488). New York: Cambridge University Press.

Pratiwi, A.I. (2011). Optimizing the Use of Youtube Video to Improve Students’ Writing Skill (A Classroom Action Research at the Second Grade of SMP Negeri 1 Juwirang Klaten in the academic Year of 2009/2010). Sebelas Maret University. Unpublished Thesis.

Pun, M. (2013). The use of multimedia technology in English language teaching: A global perspective. Crossing the border: International journal of interdisciplinary studies , 1 ISSN 2350-8752(P) volume. https://www.nepjol.info/index.php/CTBIJIS/article/view/10466 . Accessed 10 Apr 2020.

Rice, D. (2011). Appendix 1: Use of ICTs for inclusive education http://inova.snv.jussieu.fr/evenements/colloques/colloques/article.php?c=70&l=en&a=361 . Accessed 15 Mar 2020.

Robinson, J. (2019). Asian English - the British library https://www.bl.uk/british-accents-and-dialects/articles/asian-english. Accessed 15 Feb 2020 .

Sarason, Y., & Banbury, C. (2004). Active learning facilitated by using a game-show format or who doesn‘t. Journal of Management Education , 28 (4), 509–518 https://eric.ed.gov/?id=EJ797827 . Accessed 9 Feb 2020.

Schmidt, R. (1990). The Role of Consciousness in Second Language Learning. Applied Linguistics 11 , 129–158.

Schmidt, R. (2001). Attention. In P. Robinson (Ed.), Cognition and second language instruction (pp. 3–32). Cambridge: Cambridge University Press.

Schmidt, R. (2010). Attention, awareness, and individual differences in language learning . Singapore: National University of Singapore, Centre for Language Studies http://nflrc.hawaii.edu/PDFs/SCHMIDT%20Attention,%20awareness,%20and%20individual%20differences . Accessed 10 Jan 2020.

Smith, M. (1991). Speaking to many minds: On the relevance of different types of language information for the L2 learner. Second Language Research , 7 (2), 118–132 www.jstor.org/stable/43104426 . Accessed 19 Feb 2020.

Swain, M. (2005). The output hypothesis: theory and research. In E. Hinkel (Ed.), Handbook of research in second language teaching and learning (pp. 471–483). Yahweh, N.J: Lawernce EelbaumAssociates.

Swain, M., & Lapkin, S. (1995). Problems in output and the cognitive processes they generate: A step towards second language learning https://academic.oup.com/applij/article-abstract/16/3/371/184113?redirectedFrom=fulltext . Accessed 15 Mar 2020.

Tarantino, K., McDonough, J., & Hua, M. (2013). Effects of student engagement with social media on student learning: A review of literature https://www.researchgate.net/publication/280079702_Effects_of_student_engagement_with_social_media_on_student_learning_A_review_of_literature . Accessed 14 2020.

Text Inspector. (n.d.) (A tool used for measuring lexical diversity using the computations of MTLD and VOCD). https://textinspector.com/

UNESCO 2008 “UNESCO 48th International Conference on Education – Reference document” available at http://www.ibe.unesco.org/fileadmin/user_upload/Policy_Dialogue/48th_ICE/CONFINTED_48-3_English.pdf .

Vaezi, S., & Kafshgar, N. B. (2012). Learner characteristics and syntactic and lexical complexity of written products. International Journal of Linguistics, 4 (3), 671–687.

Van Gelderen, A., & Oostdam, R. (2005). Effects of fluency training on the application of linguistic operations in writing. Educational Studies in Language and Literature, 5 (2), 215–240.

Vaughan, T. (1993). Multimedia: Making it work , (1st ed., ). Osborne, Berkeley: McGraw-Hill.

Warschauer, M., & Meskill, C. (2000). Technology and Second Language Teaching. Handbook of Undergraduate Second Language Education . New Jersey: Lawrence Erilbaum.

Warschauer, M., & Grimes, D. (2007). Audience, authorship, and artifact: The emergent semiotics of web 2.0. Annual Review of Applied Linguistics , 27 , 1–23.

Winke, P., Gass, S., & Sydorenko, T. (2010). The effects of captioning videos used for foreign language listening activities Michigan State University. Language Learning & Technology. https://scholarspace.manoa.hawaii.edu/bitstream/10125/44203/1/14_01_winkegasssydorenko . Accessed 22 Jan 2020.

Wolfe-Quintero, K., Inagaki, S., & Kim, H. Y. (1998). Second language development in writing: Measures of fluency, accuracy and complexity. Technical report 17 . Manoa: University of Hawai'i Press.

Xiao, M. (2007). An empirical study of using internet-based desk videoconferencing in an EFL setting . Ohio: The College of Education of Ohio University https://etd.ohiolink.edu/pg_10?0::NO:10:P10_ACCESSION_NUM:ohiou1194703859 . Accessed 5 Jan 2020.

Yang, W. (2008). Optimization of multimedia English teaching in context creation. International Education Studies , 1 , 4 https://files.eric.ed.gov/fulltext/EJ1065458 . Accessed 5 Apr 2020.

Yesner, R. (2012). The next-generation classroom: smart, interactive and connected learning environments . IDC Government Insights.

Yoshino, S., Kano, N., & Akahori, K. (2000). The effects of English and Japanese captions on the listening comprehension of Japanese EFL students. Language Laboratory, 37, 111–130. Tokyo Institute of Technology.

Download references

Acknowledgments

This work was completed in collaboration with, and hence special thanks to, the Iraqi School, New Delhi, India. I’m also especially thankful to Mary Alyousef, as a researcher in Linguistics, for her huge language support and advice.

This research is part of an unpublished PhD dissertation at JNU, New Delhi, India and is fully funded by Al-Furat University and the Ministry of Higher Education and Scientific Research, Syria.

Author information

Authors and affiliations.

School of Languages, Literature and Culture Studies, New Delhi, India

Azzam Alobaid

You can also search for this author in PubMed   Google Scholar

Contributions

I am the sole author of this research work which has been entirely conducted by me as part of my PhD dissertation. The author read and approved the final manuscript.

Author’s information

The author of this paper, (name AA), is a teaching assistant at Al-Furat University since 2011 and currently a third year PhD student at JNU, New Delhi, India. He holds a master’s degree in TESOL (2012) and another master’s degree in Theoretical Linguistics (2016). He is an educational technologist, and his broad area of interest is technology and multimedia learning.

Corresponding author

Correspondence to Azzam Alobaid .

Ethics declarations

Ethics approval and consent to participate, consent for publication, competing interests.

The author declares that he has no competing interests.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Additional file 1., rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Alobaid, A. Smart multimedia learning of ICT: role and impact on language learners’ writing fluency— YouTube online English learning resources as an example. Smart Learn. Environ. 7 , 24 (2020). https://doi.org/10.1186/s40561-020-00134-7

Download citation

Received : 18 May 2020

Accepted : 25 August 2020

Published : 15 September 2020

DOI : https://doi.org/10.1186/s40561-020-00134-7

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Smart learning
• Multimedia materials

Research on Distributed Multimedia System in Universities Management Mode

Ieee account.

• Change Username/Password
• Update Address

Purchase Details

• Payment Options
• Order History
• View Purchased Documents

Profile Information

• Communications Preferences
• Profession and Education
• Technical Interests
• US & Canada: +1 800 678 4333
• Worldwide: +1 732 981 0060
• Contact & Support
• About IEEE Xplore
• Accessibility
• Terms of Use
• Nondiscrimination Policy
• Privacy & Opting Out of Cookies

A not-for-profit organization, IEEE is the world's largest technical professional organization dedicated to advancing technology for the benefit of humanity. © Copyright 2024 IEEE - All rights reserved. Use of this web site signifies your agreement to the terms and conditions.

Academia.edu no longer supports Internet Explorer.

To browse Academia.edu and the wider internet faster and more securely, please take a few seconds to  upgrade your browser .

Enter the email address you signed up with and we'll email you a reset link.

• We're Hiring!
• Help Center

USE OF MULTIMEDIA TO ENHANCE ORAL COMMUNICATION SKILL

Proficiency in oral communication is increasingly required both in academic and professional settings. For this reason, an increasing number of courses, taught in both public and private institutions, are addressing oral communication skills. With globalization, the number of opportunities for interactions in English has increased and so has the need to learn strategies for successful oral communication in English. Oral communication is an essential aspect of social interaction. Being able to communicate well is not only an important skill in itself, but also contributes significantly to the success of a person's personal and professional life. Speaking is used to engage in conversations, transmit information, express opinions, and contribute to discussions. The paper highlights the necessity and use of multimedia technology in an effective manner to enhance the oral communication skill. Oral presentation skill is significant for the students as it not only help them in academic but also in their professional life after formal education. The current instructional scenario in our higher educational institutes is still replete with traditional way of teaching which trained the students in reading and writing while listening and speaking skills are neglected and ignored. Thus, the teachers and students need to make integrated efforts for development of oral communication skills using modern technologies. In the present scenario multimedia is universally regarded as an essential tool in the field of teaching and research. Recently, new developments and increasing affordability of multimedia technologies have provided teachers an opportunity to teach students how to improve oral communication skills. Multimedia broadly covers the set of activities that facilitates capturing, storage, processing, transmission and display of information by electronic means.

RELATED TOPICS

•   We're Hiring!
•   Help Center
• Find new research papers in:
• Health Sciences
• Earth Sciences
• Cognitive Science
• Mathematics
• Computer Science
• Academia ©2024

International Conference on Mechatronics and Intelligent Robotics

ICMIR 2017: Recent Developments in Mechatronics and Intelligent Robotics pp 525–530 Cite as

Research on Video Transmission Technology Based on Wireless Network

• Liqiang Liu 17  
• Conference paper
• First Online: 12 November 2017

Part of the book series: Advances in Intelligent Systems and Computing ((AISC,volume 691))

With the rapid development of wireless mobile communication technology, the multimedia communication service is increasing rapidly in the mobile environment. Analyzing With the rapid development of wireless communication technology and the continuous improvement of wireless network bandwidth, the multimedia communication service is increasing rapidly in the mobile environment. The communication performance of multimedia transmission in wireless networks with limited resources, the video encoding in the main influencing factors of multimedia transmission and application is studied, through the evaluation of video transmission in wireless network quality of service by using simulation tools, such as data packet time delay, PSNR, image difference etc. Experimental results show that, when the video is transmitted over the wireless network, it is necessary to select the appropriate compress quantization parameters and GOP type according to the network condition, so as to obtain a better reconstructed video quality.

• Wireless network
• Multimedia information
• Video coding
• Network simulation

This is a preview of subscription content, log in via an institution .

Buying options

• Available as PDF
• Read on any device
• Instant download
• Own it forever
• Available as EPUB and PDF
• Compact, lightweight edition
• Dispatched in 3 to 5 business days
• Free shipping worldwide - see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

Zhou, L., Yang, Z., Wen, Y., Rodrigues, J.: Distributed wireless video scheduling with delayed control information. IEEE Trans. Circuits Syst. Video Technol. 24 (5), 889–901 (2013)

Article   Google Scholar  

Zhixing, L., Jing, Z., Li, Z.: A distributed JPEG2000 coding algorithm based on mobile agent. In: International Conference on Wireless Communications and Signal Processing, pp. 1–5 (2009)

Google Scholar  

Wang, P., Rui, D., Akyildiz, I.F.: A spatial correlation-based image compression framework for wireless multimedia sensor networks. IEEE Trans. Multimed. 13 (2), 388–401 (2011)

Gan, L., Tu, C., Liang, J., Tran, T.D., Ma, K.-K.: Undersampled boundary pre-/postfilters for low bit-rate DCT-based block coders. IEEE Trans. Image Process. 16 (2), 428–442 (2007)

Article   MathSciNet   Google Scholar  

Junejo, I., Cao, X., Foroosh, H.: Geometry of a non-overlapping multi-camera network. In: IEEE International Conference on Video and Signal Based Surveillance, p. 43. IEEE Press, Sydney (2006)

Wiegand, T., Sullivan, G., Luthra, A.: Draft ITU-T recommendation and final draft international standard of joint video specification (ITU-T Rec. H.264 ISO/IEC 14496-10 AVC). Joint Video Team (JVT) of ISO/IEC MPEG & ITU-T VCEG 8th Meeting document JVT-G050, Geneva, p. 3 (2003)

Ling, S., Xie, P., Wu, D., Xu, B.: A multi-interface multi-channel Ad Hoc routing protocol for real time video transmission. In: 2011 International Conference on Mechatronic Science Electric Engineering and Computer (MEC) (2011)

Odaka, T.: Personalized MPEG2 video-data transmission system. In: 2008 Third International Conference on Internet and Web Applications and Services, p. 6 (2008)

Download references

Acknowledgment

This research is supported by Scientific research project of Hunan Provincial Department of Education (16C0721).

Author information

Authors and affiliations.

Department of Computer, Hunan Institute of Science and Technology, Yueyang, China

Liqiang Liu

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Liqiang Liu .

Editor information

Editors and affiliations.

Shenyang Jianzhu University , Shenyang, China

SOA University , Bhubaneswar, India

Srikanta Patnaik

Montclair State University , Montclair, New Jersey, USA

Rights and permissions

Reprints and permissions

Copyright information

© 2018 Springer International Publishing AG

About this paper

Cite this paper.

Liu, L. (2018). Research on Video Transmission Technology Based on Wireless Network. In: Qiao, F., Patnaik, S., Wang, J. (eds) Recent Developments in Mechatronics and Intelligent Robotics. ICMIR 2017. Advances in Intelligent Systems and Computing, vol 691. Springer, Cham. https://doi.org/10.1007/978-3-319-70990-1_76

Download citation

DOI : https://doi.org/10.1007/978-3-319-70990-1_76

Published : 12 November 2017

Publisher Name : Springer, Cham

Print ISBN : 978-3-319-70989-5

Online ISBN : 978-3-319-70990-1

eBook Packages : Engineering Engineering (R0)

Share this paper

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Publish with us

Policies and ethics

• Find a journal
• Track your research

This paper is in the following e-collection/theme issue:

Published on 4.4.2024 in Vol 26 (2024)

Impacts of an Acute Care Telenursing Program on Discharge, Patient Experience, and Nursing Experience: Retrospective Cohort Comparison Study

Authors of this article:

Original Paper

• Courtenay R Bruce, MA, JD   ; 
• Steve Klahn, RN, MBA   ; 
• Lindsay Randle, MBA   ; 
• Xin Li, BS   ; 
• Kelkar Sayali, BS   ; 
• Barbara Johnson, BSN, MBA, DNP   ; 
• Melissa Gomez, MBA   ; 
• Meagan Howard, MHA   ; 
• Roberta Schwartz, PhD   ; 
• Farzan Sasangohar, PhD  

Houston Methodist, Houston, TX, United States

Corresponding Author:

Courtenay R Bruce, MA, JD

Houston Methodist

8100 Greenbriar Drive

Houston, TX, 77030

United States

Phone: 1 281 620 9040

Email: [email protected]

Background: Despite widespread growth of televisits and telemedicine, it is unclear how telenursing could be applied to augment nurse labor and support nursing.

Objective: This study evaluated a large-scale acute care telenurse (ACTN) program to support web-based admission and discharge processes for hospitalized patients.

Methods: A retrospective, observational cohort comparison was performed in a large academic hospital system (approximately 2100 beds) in Houston, Texas, comparing patients in our pilot units for the ACTN program (telenursing cohort) between June 15, 2022, and December 31, 2022, with patients who did not participate (nontelenursing cohort) in the same units and timeframe. We used a case mix index analysis to confirm comparable patient cases between groups. The outcomes investigated were patient experience, measured using the Hospital Consumer Assessment of Health Care Providers and Systems (HCAHCPS) survey; nursing experience, measured by a web-based questionnaire with quantitative multiple-choice and qualitative open-ended questions; time of discharge during the day (from electronic health record data); and duration of discharge education processes.

Results: Case mix index analysis found no significant case differences between cohorts ( P =.75). For the first 4 units that rolled out in phase 1, all units experienced improvement in at least 4 and up to 7 HCAHCPS domains. Scores for “communication with doctors” and “would recommend hospital” were improved significantly ( P =.03 and P =.04, respectively) in 1 unit in phase 1. The impact of telenursing in phases 2 and 3 was mixed. However, “communication with doctors” was significantly improved in 2 units ( P =.049 and P =.002), and the overall rating of the hospital and the ”would recommend hospital” scores were significantly improved in 1 unit ( P =.02 and P =04, respectively). Of 289 nurses who were invited to participate in the survey, 106 completed the nursing experience survey (response rate 106/289, 36.7%). Of the 106 nurses, 101 (95.3%) indicated that the ACTN program was very helpful or somewhat helpful to them as bedside nurses. The only noticeable difference between the telenursing and nontelenursing cohorts for the time of day discharge was a shift in the volume of patients discharged before 2 PM compared to those discharged after 2 PM at a hospital-wide level. The ACTN admissions averaged 12 minutes and 6 seconds (SD 7 min and 29 s), and the discharges averaged 14 minutes and 51 seconds (SD 8 min and 10 s). The average duration for ACTN calls was 13 minutes and 17 seconds (SD 7 min and 52 s). Traditional cohort standard practice (nontelenursing cohort) of a bedside nurse engaging in discharge and admission processes was 45 minutes, consistent with our preimplementation time study.

Conclusions: This study shows that ACTN programs are feasible and associated with improved outcomes for patient and nursing experience and reducing time allocated to admission and discharge education.

Introduction

Telemedicine, particularly video televisits, has greatly expanded in the wake of the COVID-19 pandemic [ 1 , 2 ]. Televisits have shown promise as a robust, practical, efficacious, and scalable alternative to in-person office visits that could ameliorate labor supply shortages [ 3 , 4 ]. The published evidence suggests a generally positive attitude toward televisit appointments for chronic care, focused on addressing financial and transportation barriers and improving patients’ access to care [ 5 - 7 ]. Despite the promise shown by televisits, limited attention has been paid to applying this method in the acute care setting and, in particular, on how this promising technology can be leveraged to support nurses.

Estimates suggest that approximately 200,000 open nursing positions will become available each year between 2021 and 2031 [ 8 ]. Telenursing can augment nursing labor supply, decrease nursing workload, maintain patient and nurse safety, and positively impact nursing and patient experiences [ 9 ]. However, the impact of telenursing on outcomes in acute care settings remains a research gap.

To address this gap, this study aimed to evaluate the outcomes associated with a large-scale acute care telenurse (ACTN) program to support web-based admission and discharge processes for hospitalized patients compared to patients who did not undergo the ACTN program intervention. Admission and discharge are 2 substantive and time-consuming acute care nursing tasks that involve tedious documentation in the electronic health record (EHR) and extensive interaction with patients and families to gather history and provide patient education [ 10 , 11 ]. We aimed to develop an ACTN program to augment nursing care by conducting admission and discharge processes through telenursing in a large health system. Subsequently, we discuss the impacts on 4 end points: patient experience, nursing experience, time of discharge during the day, and length of time for discharge education processes. We hypothesized that the ACTN program would be associated with higher patient experience scores and improved nursing experience compared to standard admission and discharge practices.

This study was conducted in a large academic hospital system (approximately 2100 beds) in Houston, Texas. The preimplementation methods are reported more extensively in the studies by Hehman et al [ 12 ] and Schwartz et al [ 13 ]. Program implementation was first informed by nursing time and workload surveys and pilot implementation in 4 comparatively understaffed units. The chief innovation officer, along with nursing leaders and ACTN program administrators, met with the bedside nursing staff of these 4 understaffed units to solicit their input on where and how ACTN would add value to their workflow. Bedside nursing staff provided critical input on admission processes that could be delegated to individuals working remotely with no perceived negative impact on patient experience. We conducted participatory workflow design sessions with bedside nursing staff on the ACTN program to cocreate workflow integration points where the remote team could assist [ 13 ].

Pilot Implementation and Procedures

Before implementation, the ACTN administrators trained bedside nurses in pilot units by demonstrating the use of technology during shift huddles. Then, the trainers presented slides on contact information and available support and provided a role demarcation process map, showing what the remote telenurse staff would be doing compared to what the bedside nurses needed to do to launch and conduct discharge education. Furthermore, the trainers invited the nursing staff to observe several discharges to learn how to conduct them. A software with Health Insurance Portability and Accountability Act compliance was uploaded to iPads (Apple Inc) and stored on each unit. Handheld iPads were available, and roaming iPads were made available for patients who could not hold an iPad.

The pilot implementation was staggered in a phased rollout, consisting of 3 sequenced phases, as shown in Figure 1 . Upon admission, the acute care bedside nurse contextualized the ACTN program with patients and families by handing the patient an iPad with a preloaded remote program app (Caregility) and then pressing a soft key to allow the ACTN to enter the patient’s room via the iPad screen. The ACTN introduced themselves, completed the nursing admission profile in the EHR, placed a request for a consultation, and notified the bedside nurse that the admission was completed using secure SMS text messaging [ 13 ]. A similar process was followed for discharge workflow processes, where the ACTN completed patient education on discharge instructions, confirmed the patient’s pharmacy details, confirmed discharge transportation, and arranged for departure.

Bedside nurses used their discretion regarding which patients would be appropriate for the ACTN program. They based this determination principally on whether documentation was needed and whether the patient could benefit from the undivided attention the ACTN program could afford. Furthermore, they excluded patients from the ACTN program if the patients expressed discomfort using an iPad. After the initial rollout, patients’ input was sought on their experience with the ACTN program to identify where and how improvements could be made, and this feedback was incorporated into iterative revisions in subsequent rollouts.

Pilot Outcomes Monitoring

A retrospective, observational cohort comparison was performed, in which all patients in our pilot units for the ACTN program (telenursing cohort) between June 15, 2022, and December 31, 2022, were compared with all patients who did not participate (nontelenursing cohort) in the same units in the same timeframe.

Our primary outcomes were patient experience and nursing experience. Patient experience scope was any process observable by patients [ 14 ]. We compared patient experiences in the telenursing and nontelenursing cohorts by evaluating patients’ responses to the widely used Hospital Consumer Assessment of Health Care Providers and Systems (HCAHPS) survey [ 15 ], which represented 8 aspects (called dimensions) of patient satisfaction. Each dimension was measured using a continuous variable (0 to 100 points).

For the telenursing cohort, we analyzed bedside nurses’ collective responses using a Forms (Microsoft Corp) survey conducted in April 2023. The survey consisted of 5 questions, asking them to indicate whether the ACTN program was helpful using a Likert scale with 5 items (very helpful to very unhelpful). Nurses were asked to provide open-ended comments to explain the reasons for their evaluation. At the end of the survey, we included 2 open-ended fields for nurses to describe opportunities for improvement in future rollouts and provide any additional comments.

Furthermore, we explored the time at which discharge occurred using the EHR admission, discharge, and transfer date and time. We compared the hour of the day the patient was discharged in the telenursing cohort with the hour of the day the patient was discharged in the nontelenursing cohort, hypothesizing a priori that patients might be discharged earlier in the day in the telenursing cohort. Finally, we analyzed the duration of discharge education for both cohorts, measured in minutes.

Data Analysis

The patient demographic data were available for all patients. To confirm that the telenursing cohort had similar patient demographics as the nontelenursing cohort (and therefore to confirm that nurse biases in patient selection for the ACTN program were unlikely), we conducted a case mix index (CMI) evaluation. We first isolated the population of both cohorts into adults (aged ≥18 y). We compared only those patients who were discharged home and excluded those who were on extracorporeal membrane oxygenation or those who underwent a tracheostomy. The remaining population was evaluated to determine whether there was a difference in patient acuity and severity. After confirming that patient acuity and severity were of no significant difference, we included the inpatient and observation populations to evaluate the intervention results.

For the patient experience data, independent sample t tests (2-tailed) were used to compare the telenursing and nontelenursing cohorts across different HCAHPS dimensions and units. Analysis was conducted using R software (R Foundation for Statistical Computing). For the nursing experience survey data, we used Excel (Microsoft Corp) to analyze the responses to multiple-choice, discrete questions and thematic analysis to evaluate the open-text fields. Thematic analysis allows for eliciting key themes that emerge based on recurring statements [ 16 ]. The analysis followed an inductive approach. This approach uses open-ended questions, allowing themes to emerge with a few previously articulated assumptions on responses. Given the limited content, CRB served as the primary coder. Coding labels were used for data contextualizing, allowing for new themes to emerge throughout the coding process, using a codebook [ 16 , 17 ]. We stored emergent patterns and themes in an electronic format.

Ethical Considerations

The hospital’s review board determined that the ACTN pilot would not be considered regulated human subjects research. All data reported in this study were aggregated and deidentified.

The demographics of the telenursing and nontelenursing cohorts were relatively similar. Both cohorts had an average age of 60 years with an SD of 16.91; had a similar distribution in race and ethnicity (approximately 92/2319, 3.96% Asian; 525/2319, 22.64% Black; 425/2319, 18.33% Hispanic; 70/2319, 3.02% Native American, declined to identify, or other categories; and 1202/2319, 51.83% White); and had a similar distribution in female participants versus male participants (1249/2319, 53.86% vs 1070/2319, 46.14%). To further understand the population, the CMI analysis for acuity and severity showed that the CMI was slightly higher in the telenursing cohort than in the nontelenursing cohort, but the difference was not statistically significant ( P =.75).

Patient Experience

Among the first 4 units that rolled out in phase 1, all units experienced improvement in at least 4 and up to 7 HCAHPS domains (Table S1 in Multimedia Appendix 1 ). On average, 6 out of 8 HCAHPS domains were improved for patients in the telenursing cohort. All 4 units experienced improvements in the “overall rating” domain, and 3 of the 4 units experienced improvements in “likelihood to recommend” domain for patients in the telenursing cohort compared to those in the nontelenursing cohort within the same units. The improvement scores ranged from 1.4% for the neurosurgery unit (36 beds) to 11.6% for the medical unit (37 beds). Furthermore, all 4 units in the first phase of roll out experienced improved scores in the “responsiveness” domain by >4 points (ranging from 5% to 10.1%). A total of 2 out of the 4 units also experienced improvements in the “communication with nurses” (ranging from 1.7% to 3%) and “communication about medicines” (ranging from 3.3% to 11.7%) domains. The 2 units that did not experience improvement in the communication domains were the combined medical and surgery neurology and neurosurgical units (36 beds). Only the neurosurgical unit showed statistically significant improvements in 2 dimensions: “communication with doctors” ( P =.03) and “would recommend hospital” ( P =.04).

For the 7 units that rolled out during phase 2, only 1 orthopedic surgery unit (28 beds) experienced improvements in every domain (ranging from 0.9% to 12.5%). Medical observation unit 1 also improved in 5 areas. However, only improvements in “communication with doctors” ( P =.002), “overall rating of hospital” ( P =.02), and “would recommend hospital” ( P =.04) were statistically significant . The remaining units experienced improvements in some domains for the telenursing cohort compared to the nontelenursing cohort, with no improvement in other domains. However, the scores for “communication with nurses” and “communication with doctors” domains were improved for most of the units that rolled out in phase 2 (Table S2 in Multimedia Appendix 1 ).

For the 2 units that rolled out in phase 3, both of which were surgical cardiac units with 36 beds, 1 unit experienced improvement in every domain except “responsiveness” (ranging from 1% to 12%). The other unit only experienced improvement in the “communication with doctors” (4.9%) and “care transitions” domains (1.1%). However, none of these improvements were statistically significant (Table S3 in Multimedia Appendix 1 ).

Nursing Experience

Of the 289 nurses who were invited to participate in the survey, 106 completed the survey (36.7% response rate). Of the 106 nurses, 101 (95.3%) indicated that the ACTN program was “very helpful” or “somewhat helpful” to them as bedside nurses.

Quantitative Findings

The main reasons nurses gave for the program’s helpfulness included that it saved them time (94/106, 88.7%), allowed them to focus on more urgent clinical needs (90/106, 84.9%), allowed them to focus on activities they felt were more in line with their skill level (55/106, 51.9%), and allowed patients to have undivided attention for their discharge education (52/106, 49.1%). Among the 5 nurses who indicated that the ACTN program was somewhat unhelpful or very unhelpful, 3 (60%) indicated that workflows were not clear or needed further refinement or clarification. Furthermore, the nurse respondents shared several barriers and provided opportunities for improvement, with 91 (85.8%) out of 106 nurses offering suggestions.

Qualitative Findings

For the free-text explanation fields, all but 3 nurses (103/106, 97.2%) provided additional comments on the ACTN program helpfulness. Three themes emerged from the qualitative analysis of the free-text comments: (1) most of the nurses’ comments reflected that telenurses help bedside nurses save time, (2) respondents indicated that extra hands provided emotional and physical support in providing patient care, and (3) respondents perceived an improvement in patient safety by having a telenurse who could “catch missed” issues.

Time Saving

One of the perceived benefits of the telenursing program was saving time. One nurse said the following:

... Just putting in home medications alone takes up so much time. This new telenurse service helps [save time]

Several nurses highlighted that admission and discharge processes are so complex and time-consuming that shifting this work to the ACTN program freed nurses to perform other activities, as reflected by this nurse:

The tele RN is able to spend as much time possible sufficiently educating an admission or discharge while allowing me time to respond to the needs of my other patients saving me time on one patient especially charting.

Emotional and Physical Support

For the second theme, several responses focused less on time management and perceived efficiencies and instead centered more on the emotional appeal and support in having an extra hand, as one nurse mentioned:

Being in such a fast-paced unit, it can be a bit stressful with so many discharges and admissions. Having a helpful hand is beneficial.

Improved Patient Safety

Finally, the third theme was perceived improvement in patient safety by having a telenurse who could “catch missed” issues (eg, an incorrectly identified pharmacy details), simultaneously allowing the primary bedside nurse to focus more intensely on other needs, essentially creating a 2-fold safety promotion. Some nurses noted that they could begin carrying out orders while the telenurses began completing the admission, facilitating quicker treatment and resolution of care needs, thereby improving the safety and quality of care. One nurse mentioned the following:

Allows [telenurses] to take on thorough and accurate admissions, while also preventing any rushing the patient might experience from the primary RN.

When asked for areas of improvement, the most recurring theme was having 24 hours of support during the weekend and during the week. The second theme for improvement was the reduced time to connect to a telenurse. The third theme was the availability of iPads. Nurses mentioned that iPads could sometimes be unavailable in patients’ rooms or they may not be fully charged.

Time of Discharge

The time of day distribution is presented in Figure 2 . The only noticeable difference between the telenursing and nontelenursing cohorts was a shift in the volume of patients discharged before 2 PM compared with those discharged after 2 PM at a hospital-wide level ( Table 1 ). At an individual unit level, these results were not consistent and could be further explored by patient population and their needs to discharge. The variation was further illustrated when reviewing the length of stay of patients in the telenursing and nontelenursing cohorts. Only 5 out of the 12 units showed a decrease in the average inpatient length of stay.

Discharge Length

The ACTN admissions averaged 12 minutes and 6 seconds (SD 7 min and 29 s), and the discharges averaged 14 minutes and 51 seconds (SD 8 min and 10 s). The average duration for ACTN calls was 13 minutes and 17 seconds (SD 7 min and 52 s). Traditional cohort standard practice of a bedside nurse engaging in discharge and admission processes was 45 minutes, consistent with our preimplementation nursing time study.

Principal Findings

Our results suggest that the ACTN program was associated with positive nursing experiences because it saved time. Furthermore, the ACTN program was associated with higher HCAHPS scores in several domains but only in the first series of units that piloted the intervention. In phase 1, the improvement in “communication with doctors” and “would recommend hospital” scores in 1 unit was statistically significant. In phase 2, the improvement in “communication with doctors” score was significant in 2 units and that in “overall rating of hospital” and “would recommend hospital” scores were significant in 1 unit. The time of day discharge was nearly the same in both the telenursing and nontelenursing cohorts. The duration for discharge processes was less than half in the ACTN cohort compared to the nonintervention cohort.

At the time of writing this paper, the United States was experiencing a critical nursing shortage that will likely reach an epidemic level in the next few decades [ 8 ]. Despite the promise shown by telenursing, to our knowledge, only 1 existing paper documents the impact of ACTN programs on HCAHPS-measured patient satisfaction using a small cohort of patients in a single, time-limited pre- and posttelenursing analysis [ 18 ]. A study by Schuelke et al [ 18 ] revealed a 6.2% increase in “communication with meds” and 12.7% increase in “communication with nursing” domain scores; other HCAHPS domains were not evaluated. This research builds upon the promising work of Schuelke et al [ 18 ], evaluating the impact of an ACTN program on several units with a much larger cohort of patients using a staggered rollout and comparing all HCAHPS domains between telenursing and nontelenursing cohorts within the same time frame and in the same units.

By conducting granular HCAHPS analyses, we identified what we believed to be a time sequence variability in that units that rolled out in phase 1 performed considerably stronger in HCAHPS impacts than units that rolled out in later phases. An explanation for this sequence effect might be that some later adopters had less potential for high effect size, given that the first 4 units of the rollout were specifically chosen for their staffing problems compared to later units. ACTN support might have augmented the staffing support to such a degree that allowed the impacts of the program to be more salient. An alternative explanation is that the early adopters and promoters tend to have greater diffusion uptake, greater saturation and adoptability, and greater impacts compared to late adopters or those resistant to adoption [ 19 , 20 ]. Our anecdotal evidence suggests that early adopters might have wanted the telenursing program to succeed; therefore, they applied consistent implementation practices to ensure success. Adopters in later stages were more aware of barriers and potential downsides and might have been more ambivalent about telenursing and, therefore, less likely to modify their behaviors to promote the telenursing program’s success.

Another interesting finding was that the ACTN program seemed to be effective for both medical and surgical units of all specialties. Phase 1 was a mix of medical and surgical units; however, all units experienced increases in scores. Phases 2 and 3 experienced mixed results, without a clear lead for one specialty over the other. This may suggest that ACTN programs are broadly applicable across acute settings and that success depends most crucially on the need and desire of unit leaders.

Our time of day discharge findings showed only a few quantitative positive efficiencies. However, our discharge duration analysis and nursing experience survey results showed that ACTN has major time-saving benefits for nurses, suggesting a discrepancy between perceived and actual time savings versus time-of-day discharge savings. One explanation for this discrepancy may be that many factors beyond nursing impact the time of the day a patient is discharged; therefore, while the bedside nurses’ time is saved, the remaining discharge processes beyond nurses remain unaffected. Specifically, there are 3 segments of time during discharge processes: (1) the time for the discharge order and medication reconciliation [ 21 ] to the time the after-visit summary (AVS) is populated and printed [ 22 ]; (2) the time the AVS is completed and printed to the time the discharge instructions are provided; and (3) the time from providing the discharge instructions to the actual discharge ( Figure 3 ). Notably, telenurses’ involvement is currently limited to only the second segment of time. Specifically, telenurses’ involvement is not initiated until the AVS is printed by the nurse, which means that telenurses cannot positively impact any discharge activity that occurs between the time the discharge order is written and the time the AVS is printed. However, there are inefficiencies and bottlenecks in discharge processes that occur well before the AVS is printed [ 23 , 24 ]. For instance, the discharging physician may write a conditional discharge order early in the morning, listing conditions that cannot be fulfilled within a few hours or it may take bedside nursing longer than anticipated time to print the AVS.

To create a wider cascade effect for positively impacting the discharge processes for all segments of time, we are currently trying to obtain greater transparency through EHR reporting in what occurs for segments 1 and 3. For instance, at present, we know that at least 2 hospitals in our 8-hospital system have high incidence rates of conditional discharge orders that should be reduced. One hospital anecdotally reports that the discharging physician identifies incorrect pharmacies, which requires a nurse to send the scripts back to the discharging pharmacist to reconcile before discharge education can occur [ 25 ]; however, the prevalence and location of these issues remain speculative. Segment 3 is a black box of time [ 26 ]—the time it takes for hospital transport or an ambulance to arrive and move the patient to their destination and the time it takes for the family to pick up the patient. All these factors impact the discharge processes and need to be fully elucidated, explored, and streamlined. Furthermore, we hope to facilitate processes that enable telenurses to print the AVS, to remove the dependency on bedside nurses to begin the discharge education process.

Limitations

This study has several noteworthy limitations. First, the study was conducted in 1 health system and the results may not be generalizable to other settings with different patient populations, processes, and implementation strategies [ 27 ]. Second, in this study, we did not control for other factors that could impact patient and provider satisfaction as well as discharge times; telenursing can only improve upon one component in a complex set of factors limiting discharge efficiency and satisfaction outcomes. Finally, participating nurses were aware of the ongoing study, and this knowledge might have affected their behavior [ 28 ].

Future Directions

After the completion of this pilot study, the ACTN admission and discharge program has been rolled out to pilot medical units and all surgical and observation units. Our rationale for expansion rested on the premise that nursing experience is important to maintain and strengthen, particularly at a time when turnover is high in the health care industry in general. It is important to reduce staff inefficiencies in workload as a means of preserving or strengthening organizational morale and cost saving. Because our nursing experience findings for the ACTN program heavily supported the program, this served as the primary motivation for expansion. The nursing experience findings, coupled with the findings related to time-savings in discharge education and modest improvement, though not negative, in the HCAHPS findings for the ACTN program compared to the nontelenursing cohort, further supported expansion.

The initial scope for expansion included a complete system-wide implementation for all admissions and discharges. Furthermore, we are planning to expand the ACTN program beyond admissions and discharges. Responsive to qualitative feedback reported earlier, the next phase of the ACTN program will add safeguards on high-risk medications by having telenurses conduct double-checks, skin assessments, hourly rounding assistance, and auditing of safety functions and educational activities. These activities were chosen because they are time-intensive for nursing staff on the patient floors. Additional support in these areas would be a staff morale booster in addition to improved efficiencies for bedside nursing. Conducting hourly rounding using the ACTN program will require more time and resources; however, conducting high-quality, uninterrupted hourly rounds is known to be effective at improving patient safety and patient experience outcomes [ 29 ]. Therefore, we suspect that the ACTN program will have some positive impacts if rounds are consistently conducted, even if conducted virtually.

In addition, the ACTNs have been motivating other specialties to adopt or consider a similar program as the ACTN program to support stretched staffing. These specialties include respiratory care, in which virtual support can quickly identify patients in need of intensive on-site support; pharmacy, in which direct communication with staff on medications and patient training can happen through virtual means; infection control, in which room environments can be reviewed through virtual audits, moving quickly from floor to floor; and guest relations and spiritual care, in which patients can be visited virtually upon patient request. Furthermore, physicians who wish to either virtually enter inpatient rooms during their clinic days or from home can quickly drop in to see patients using the virtual program. For these groups to further develop advanced inpatient telemedicine programs, additional technology will be required, including cameras that can zoom into various portions of the room and advanced sound capabilities. Future work could expand programs similar to ACTN to specialties such as respiratory therapy, pharmacy, infection prevention, and spiritual care.

Conclusions

This study provides preliminary evidence suggesting that telenursing may effectively address nursing shortages in acute care settings and positively impact patient and provider satisfaction as well as admission and discharge times. More work is needed to validate the findings in other settings, use other satisfaction metrics, and investigate the impact of telenursing on the quality of care and cost.

Acknowledgments

The authors would like to thank Jacob M Kolman, MA, ISMPP CMPP, senior scientific writer, Houston Methodist Academic Institute, for the critical review and for providing formatting feedback on this manuscript. The authors would also like to thank Amir Hossein Javid for his help with statistical analysis.

Data Availability

Data sharing is not applicable as no data sets were generated during this study.

Authors' Contributions

All authors were involved in the conceptualization, review and approval, and writing of the manuscript. LR, BJ, MG, RS, SK, and MH were extensively involved in the implementation of the project. BJ, MH, SK, and MG conducted the training. SK and XL conducted the analyses. CRB wrote and edited the manuscript, inserted and refined the citations, and provided critical feedback during implementation and analyses. CRB and FS were involved in all stages of writing and publication. All authors meaningfully contributed to the drafting, writing, brainstorming, executing, finalizing, and approving of the manuscript.

Conflicts of Interest

None declared.

Additional outcome information for Hospital Consumer Assessment of Health Care Providers and Systems, time of day discharges, and discharge education processes.

• Greiwe J. Telemedicine lessons learned during the COVID-19 pandemic. Curr Allergy Asthma Rep. Jan 21, 2022;22(1):1-5. [ FREE Full text ] [ CrossRef ] [ Medline ]
• Doraiswamy S, Abraham A, Mamtani R, Cheema S. Use of telehealth during the COVID-19 pandemic: scoping review. J Med Internet Res. Dec 01, 2020;22(12):e24087. [ FREE Full text ] [ CrossRef ] [ Medline ]
• Ekeland AG, Bowes A, Flottorp S. Effectiveness of telemedicine: a systematic review of reviews. Int J Med Inform. Nov 2010;79(11):736-771. [ CrossRef ] [ Medline ]
• Nanda M, Sharma R. A review of patient satisfaction and experience with telemedicine: a virtual solution during and beyond COVID-19 pandemic. Telemed J E Health. Dec 01, 2021;27(12):1325-1331. [ CrossRef ] [ Medline ]
• Yu X, Bayram A. Managing capacity for virtual and office appointments in chronic care. Health Care Manag Sci. Dec 2021;24(4):742-767. [ FREE Full text ] [ CrossRef ] [ Medline ]
• Laskowski ER, Johnson SE, Shelerud RA, Lee JA, Rabatin AE, Driscoll SW, et al. The telemedicine musculoskeletal examination. Mayo Clin Proc. Aug 2020;95(8):1715-1731. [ FREE Full text ] [ CrossRef ] [ Medline ]
• Hughes L, Petrella A, Phillips N, Taylor RM. Virtual care and the impact of COVID-19 on nursing: a single centre evaluation. J Adv Nurs. Feb 2022;78(2):498-509. [ FREE Full text ] [ CrossRef ] [ Medline ]
• Pulse on the nation’s nurses survey series: COVID-19 two-year impact assessment survey. American Nurses Foundation. Mar 1, 2022. URL: https:/​/www.​nursingworld.org/​~4a2260/​contentassets/​872ebb13c63f44f6b11a1bd0c74907c9/​covid-19-two-year-impact-assessment-written-report-final.​pdf [accessed 2023-04-14]
• Abraham C, Jensen C, Rossiter L, Dittman Hale D. Telenursing and remote patient monitoring in cardiovascular health. Telemed J E Health. Sep 08, 2023. (forthcoming). [ CrossRef ] [ Medline ]
• Rouleau G, Gagnon M, Côté J, Payne-Gagnon J, Hudson E, Dubois CA. Impact of information and communication technologies on nursing care: results of an overview of systematic reviews. J Med Internet Res. Apr 25, 2017;19(4):e122. [ FREE Full text ] [ CrossRef ] [ Medline ]
• Moy AJ, Schwartz JM, Chen R, Sadri S, Lucas E, Cato KD, et al. Measurement of clinical documentation burden among physicians and nurses using electronic health records: a scoping review. J Am Med Inform Assoc. Apr 23, 2021;28(5):998-1008. [ FREE Full text ] [ CrossRef ] [ Medline ]
• Hehman MC, Fontenot NM, Drake GK, Musgrove RS. Leveraging digital technology in nursing. Health Emerg Disaster Nurs. 2023;10:41-45. [ CrossRef ]
• Schwartz RL, Hamlin SK, Vozzella GM, Randle LN, Klahn S, Maris GJ, et al. Utilizing telenursing to supplement acute care nursing in an era of workforce shortages: a feasibility pilot. Comput Inform Nurs. Feb 01, 2024;42(2):151-157. [ CrossRef ] [ Medline ]
• Cleary PD. Evolving concepts of patient-centered care and the assessment of patient care experiences: optimism and opposition. J Health Polit Policy Law. Aug 28, 2016;41(4):675-696. [ CrossRef ] [ Medline ]
• CAHPS hospital survey. Agency for Healthcare Research and Quality. URL: https://www.ahrq.gov/cahps/surveys-guidance/hospital/index.html [accessed 2023-10-12]
• Strauss A, Corbin JM. Basics of Qualitative Research: Grounded Theory Procedures and Techniques. Thousand Oaks, CA. Sage Publications; 1990.
• Hennink MM, Kaiser BN, Marconi VC. Code saturation versus meaning saturation: how many interviews are enough? Qual Health Res. Mar 2017;27(4):591-608. [ FREE Full text ] [ CrossRef ] [ Medline ]
• Schuelke S, Aurit S, Connot N, Denney S. Virtual nursing: the new reality in quality care. Nurs Adm Q. 2019;43(4):322-328. [ CrossRef ] [ Medline ]
• Rogers E. Diffusion of Innovations. New York, NY. Free Press of Glenco; 1962.
• Mohammadi MM, Poursaberi R, Salahshoor MR. Evaluating the adoption of evidence-based practice using Rogers's diffusion of innovation theory: a model testing study. Health Promot Perspect. 2018;8(1):25-32. [ FREE Full text ] [ CrossRef ] [ Medline ]
• Fernandes BD, Almeida PH, Foppa AA, Sousa CT, Ayres LR, Chemello C. Pharmacist-led medication reconciliation at patient discharge: a scoping review. Res Social Adm Pharm. May 2020;16(5):605-613. [ CrossRef ] [ Medline ]
• Federman A, Sarzynski E, Brach C, Francaviglia P, Jacques J, Jandorf L, et al. Challenges optimizing the after visit summary. Int J Med Inform. Dec 2018;120:14-19. [ FREE Full text ] [ CrossRef ] [ Medline ]
• Bristol AA, Elmore CE, Weiss ME, Barry LA, Iacob E, Johnson EP, et al. Mixed-methods study examining family carers' perceptions of the relationship between intrahospital transitions and patient readiness for discharge. BMJ Qual Saf. Aug 2023;32(8):447-456. [ FREE Full text ] [ CrossRef ] [ Medline ]
• Piniella NR, Fuller TE, Smith L, Salmasian H, Yoon CS, Lipsitz SR, et al. Early expected discharge date accuracy during hospitalization: a multivariable analysis. J Med Syst. May 12, 2023;47(1):63. [ FREE Full text ] [ CrossRef ] [ Medline ]
• Schnipper JL, Reyes Nieva H, Yoon C, Mallouk M, Mixon AS, Rennke S, et al. What works in medication reconciliation: an on-treatment and site analysis of the MARQUIS2 study. BMJ Qual Saf. Aug 2023;32(8):457-469. [ CrossRef ] [ Medline ]
• Abu HO, Anatchkova MD, Erskine NA, Lewis J, McManus DD, Kiefe CI, et al. Are we "missing the big picture" in transitions of care? Perspectives of healthcare providers managing patients with unplanned hospitalization. Appl Nurs Res. Dec 2018;44:60-66. [ FREE Full text ] [ CrossRef ] [ Medline ]
• Leung L. Validity, reliability, and generalizability in qualitative research. J Family Med Prim Care. 2015;4(3):324-327. [ FREE Full text ] [ CrossRef ] [ Medline ]
• Merrett F. Reflections on the Hawthorne effect. Educ Psychol. Jan 2006;26(1):143-146. [ CrossRef ]
• Mitchell M, Lavenberg JG, Trotta RL, Umscheid CA. Hourly rounding to improve nursing responsiveness: a systematic review. J Nurs Adm. Sep 2014;44(9):462-472. [ FREE Full text ] [ CrossRef ] [ Medline ]

Abbreviations

Edited by T de Azevedo Cardoso, G Eysenbach; submitted 06.11.23; peer-reviewed by C Jensen; comments to author 08.12.23; revised version received 16.01.24; accepted 17.02.24; published 04.04.24.

©Courtenay R Bruce, Steve Klahn, Lindsay Randle, Xin Li, Kelkar Sayali, Barbara Johnson, Melissa Gomez, Meagan Howard, Roberta Schwartz, Farzan Sasangohar. Originally published in the Journal of Medical Internet Research (https://www.jmir.org), 04.04.2024.

This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• My Account Login
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 26 March 2024

Predicting and improving complex beer flavor through machine learning

• Michiel Schreurs   ORCID: orcid.org/0000-0002-9449-5619 1 , 2 , 3   na1 ,
• Supinya Piampongsant 1 , 2 , 3   na1 ,
• Miguel Roncoroni   ORCID: orcid.org/0000-0001-7461-1427 1 , 2 , 3   na1 ,
• Lloyd Cool   ORCID: orcid.org/0000-0001-9936-3124 1 , 2 , 3 , 4 ,
• Beatriz Herrera-Malaver   ORCID: orcid.org/0000-0002-5096-9974 1 , 2 , 3 ,
• Christophe Vanderaa   ORCID: orcid.org/0000-0001-7443-5427 4 ,
• Florian A. Theßeling 1 , 2 , 3 ,
• Łukasz Kreft   ORCID: orcid.org/0000-0001-7620-4657 5 ,
• Alexander Botzki   ORCID: orcid.org/0000-0001-6691-4233 5 ,
• Philippe Malcorps 6 ,
• Luk Daenen 6 ,
• Tom Wenseleers   ORCID: orcid.org/0000-0002-1434-861X 4 &
• Kevin J. Verstrepen   ORCID: orcid.org/0000-0002-3077-6219 1 , 2 , 3  

Nature Communications volume  15 , Article number:  2368 ( 2024 ) Cite this article

48k Accesses

846 Altmetric

Metrics details

• Chemical engineering
• Gas chromatography
• Machine learning
• Metabolomics
• Taste receptors

The perception and appreciation of food flavor depends on many interacting chemical compounds and external factors, and therefore proves challenging to understand and predict. Here, we combine extensive chemical and sensory analyses of 250 different beers to train machine learning models that allow predicting flavor and consumer appreciation. For each beer, we measure over 200 chemical properties, perform quantitative descriptive sensory analysis with a trained tasting panel and map data from over 180,000 consumer reviews to train 10 different machine learning models. The best-performing algorithm, Gradient Boosting, yields models that significantly outperform predictions based on conventional statistics and accurately predict complex food features and consumer appreciation from chemical profiles. Model dissection allows identifying specific and unexpected compounds as drivers of beer flavor and appreciation. Adding these compounds results in variants of commercial alcoholic and non-alcoholic beers with improved consumer appreciation. Together, our study reveals how big data and machine learning uncover complex links between food chemistry, flavor and consumer perception, and lays the foundation to develop novel, tailored foods with superior flavors.

Similar content being viewed by others

BitterSweet: Building machine learning models for predicting the bitter and sweet taste of small molecules

Rudraksh Tuwani, Somin Wadhwa & Ganesh Bagler

Sensory lexicon and aroma volatiles analysis of brewing malt

Xiaoxia Su, Miao Yu, … Tianyi Du

Predicting odor from molecular structure: a multi-label classification approach

Kushagra Saini & Venkatnarayan Ramanathan

Introduction

Predicting and understanding food perception and appreciation is one of the major challenges in food science. Accurate modeling of food flavor and appreciation could yield important opportunities for both producers and consumers, including quality control, product fingerprinting, counterfeit detection, spoilage detection, and the development of new products and product combinations (food pairing) 1 , 2 , 3 , 4 , 5 , 6 . Accurate models for flavor and consumer appreciation would contribute greatly to our scientific understanding of how humans perceive and appreciate flavor. Moreover, accurate predictive models would also facilitate and standardize existing food assessment methods and could supplement or replace assessments by trained and consumer tasting panels, which are variable, expensive and time-consuming 7 , 8 , 9 . Lastly, apart from providing objective, quantitative, accurate and contextual information that can help producers, models can also guide consumers in understanding their personal preferences 10 .

Despite the myriad of applications, predicting food flavor and appreciation from its chemical properties remains a largely elusive goal in sensory science, especially for complex food and beverages 11 , 12 . A key obstacle is the immense number of flavor-active chemicals underlying food flavor. Flavor compounds can vary widely in chemical structure and concentration, making them technically challenging and labor-intensive to quantify, even in the face of innovations in metabolomics, such as non-targeted metabolic fingerprinting 13 , 14 . Moreover, sensory analysis is perhaps even more complicated. Flavor perception is highly complex, resulting from hundreds of different molecules interacting at the physiochemical and sensorial level. Sensory perception is often non-linear, characterized by complex and concentration-dependent synergistic and antagonistic effects 15 , 16 , 17 , 18 , 19 , 20 , 21 that are further convoluted by the genetics, environment, culture and psychology of consumers 22 , 23 , 24 . Perceived flavor is therefore difficult to measure, with problems of sensitivity, accuracy, and reproducibility that can only be resolved by gathering sufficiently large datasets 25 . Trained tasting panels are considered the prime source of quality sensory data, but require meticulous training, are low throughput and high cost. Public databases containing consumer reviews of food products could provide a valuable alternative, especially for studying appreciation scores, which do not require formal training 25 . Public databases offer the advantage of amassing large amounts of data, increasing the statistical power to identify potential drivers of appreciation. However, public datasets suffer from biases, including a bias in the volunteers that contribute to the database, as well as confounding factors such as price, cult status and psychological conformity towards previous ratings of the product.

Classical multivariate statistics and machine learning methods have been used to predict flavor of specific compounds by, for example, linking structural properties of a compound to its potential biological activities or linking concentrations of specific compounds to sensory profiles 1 , 26 . Importantly, most previous studies focused on predicting organoleptic properties of single compounds (often based on their chemical structure) 27 , 28 , 29 , 30 , 31 , 32 , 33 , thus ignoring the fact that these compounds are present in a complex matrix in food or beverages and excluding complex interactions between compounds. Moreover, the classical statistics commonly used in sensory science 34 , 35 , 36 , 37 , 38 , 39 require a large sample size and sufficient variance amongst predictors to create accurate models. They are not fit for studying an extensive set of hundreds of interacting flavor compounds, since they are sensitive to outliers, have a high tendency to overfit and are less suited for non-linear and discontinuous relationships 40 .

In this study, we combine extensive chemical analyses and sensory data of a set of different commercial beers with machine learning approaches to develop models that predict taste, smell, mouthfeel and appreciation from compound concentrations. Beer is particularly suited to model the relationship between chemistry, flavor and appreciation. First, beer is a complex product, consisting of thousands of flavor compounds that partake in complex sensory interactions 41 , 42 , 43 . This chemical diversity arises from the raw materials (malt, yeast, hops, water and spices) and biochemical conversions during the brewing process (kilning, mashing, boiling, fermentation, maturation and aging) 44 , 45 . Second, the advent of the internet saw beer consumers embrace online review platforms, such as RateBeer (ZX Ventures, Anheuser-Busch InBev SA/NV) and BeerAdvocate (Next Glass, inc.). In this way, the beer community provides massive data sets of beer flavor and appreciation scores, creating extraordinarily large sensory databases to complement the analyses of our professional sensory panel. Specifically, we characterize over 200 chemical properties of 250 commercial beers, spread across 22 beer styles, and link these to the descriptive sensory profiling data of a 16-person in-house trained tasting panel and data acquired from over 180,000 public consumer reviews. These unique and extensive datasets enable us to train a suite of machine learning models to predict flavor and appreciation from a beer’s chemical profile. Dissection of the best-performing models allows us to pinpoint specific compounds as potential drivers of beer flavor and appreciation. Follow-up experiments confirm the importance of these compounds and ultimately allow us to significantly improve the flavor and appreciation of selected commercial beers. Together, our study represents a significant step towards understanding complex flavors and reinforces the value of machine learning to develop and refine complex foods. In this way, it represents a stepping stone for further computer-aided food engineering applications 46 .

To generate a comprehensive dataset on beer flavor, we selected 250 commercial Belgian beers across 22 different beer styles (Supplementary Fig.  S1 ). Beers with ≤ 4.2% alcohol by volume (ABV) were classified as non-alcoholic and low-alcoholic. Blonds and Tripels constitute a significant portion of the dataset (12.4% and 11.2%, respectively) reflecting their presence on the Belgian beer market and the heterogeneity of beers within these styles. By contrast, lager beers are less diverse and dominated by a handful of brands. Rare styles such as Brut or Faro make up only a small fraction of the dataset (2% and 1%, respectively) because fewer of these beers are produced and because they are dominated by distinct characteristics in terms of flavor and chemical composition.

Extensive analysis identifies relationships between chemical compounds in beer

For each beer, we measured 226 different chemical properties, including common brewing parameters such as alcohol content, iso-alpha acids, pH, sugar concentration 47 , and over 200 flavor compounds (Methods, Supplementary Table  S1 ). A large portion (37.2%) are terpenoids arising from hopping, responsible for herbal and fruity flavors 16 , 48 . A second major category are yeast metabolites, such as esters and alcohols, that result in fruity and solvent notes 48 , 49 , 50 . Other measured compounds are primarily derived from malt, or other microbes such as non- Saccharomyces yeasts and bacteria (‘wild flora’). Compounds that arise from spices or staling are labeled under ‘Others’. Five attributes (caloric value, total acids and total ester, hop aroma and sulfur compounds) are calculated from multiple individually measured compounds.

As a first step in identifying relationships between chemical properties, we determined correlations between the concentrations of the compounds (Fig.  1 , upper panel, Supplementary Data  1 and 2 , and Supplementary Fig.  S2 . For the sake of clarity, only a subset of the measured compounds is shown in Fig.  1 ). Compounds of the same origin typically show a positive correlation, while absence of correlation hints at parameters varying independently. For example, the hop aroma compounds citronellol, and alpha-terpineol show moderate correlations with each other (Spearman’s rho=0.39 and 0.57), but not with the bittering hop component iso-alpha acids (Spearman’s rho=0.16 and −0.07). This illustrates how brewers can independently modify hop aroma and bitterness by selecting hop varieties and dosage time. If hops are added early in the boiling phase, chemical conversions increase bitterness while aromas evaporate, conversely, late addition of hops preserves aroma but limits bitterness 51 . Similarly, hop-derived iso-alpha acids show a strong anti-correlation with lactic acid and acetic acid, likely reflecting growth inhibition of lactic acid and acetic acid bacteria, or the consequent use of fewer hops in sour beer styles, such as West Flanders ales and Fruit beers, that rely on these bacteria for their distinct flavors 52 . Finally, yeast-derived esters (ethyl acetate, ethyl decanoate, ethyl hexanoate, ethyl octanoate) and alcohols (ethanol, isoamyl alcohol, isobutanol, and glycerol), correlate with Spearman coefficients above 0.5, suggesting that these secondary metabolites are correlated with the yeast genetic background and/or fermentation parameters and may be difficult to influence individually, although the choice of yeast strain may offer some control 53 .

Spearman rank correlations are shown. Descriptors are grouped according to their origin (malt (blue), hops (green), yeast (red), wild flora (yellow), Others (black)), and sensory aspect (aroma, taste, palate, and overall appreciation). Please note that for the chemical compounds, for the sake of clarity, only a subset of the total number of measured compounds is shown, with an emphasis on the key compounds for each source. For more details, see the main text and Methods section. Chemical data can be found in Supplementary Data  1 , correlations between all chemical compounds are depicted in Supplementary Fig.  S2 and correlation values can be found in Supplementary Data  2 . See Supplementary Data  4 for sensory panel assessments and Supplementary Data  5 for correlation values between all sensory descriptors.

Interestingly, different beer styles show distinct patterns for some flavor compounds (Supplementary Fig.  S3 ). These observations agree with expectations for key beer styles, and serve as a control for our measurements. For instance, Stouts generally show high values for color (darker), while hoppy beers contain elevated levels of iso-alpha acids, compounds associated with bitter hop taste. Acetic and lactic acid are not prevalent in most beers, with notable exceptions such as Kriek, Lambic, Faro, West Flanders ales and Flanders Old Brown, which use acid-producing bacteria ( Lactobacillus and Pediococcus ) or unconventional yeast ( Brettanomyces ) 54 , 55 . Glycerol, ethanol and esters show similar distributions across all beer styles, reflecting their common origin as products of yeast metabolism during fermentation 45 , 53 . Finally, low/no-alcohol beers contain low concentrations of glycerol and esters. This is in line with the production process for most of the low/no-alcohol beers in our dataset, which are produced through limiting fermentation or by stripping away alcohol via evaporation or dialysis, with both methods having the unintended side-effect of reducing the amount of flavor compounds in the final beer 56 , 57 .

Besides expected associations, our data also reveals less trivial associations between beer styles and specific parameters. For example, geraniol and citronellol, two monoterpenoids responsible for citrus, floral and rose flavors and characteristic of Citra hops, are found in relatively high amounts in Christmas, Saison, and Brett/co-fermented beers, where they may originate from terpenoid-rich spices such as coriander seeds instead of hops 58 .

Tasting panel assessments reveal sensorial relationships in beer

To assess the sensory profile of each beer, a trained tasting panel evaluated each of the 250 beers for 50 sensory attributes, including different hop, malt and yeast flavors, off-flavors and spices. Panelists used a tasting sheet (Supplementary Data  3 ) to score the different attributes. Panel consistency was evaluated by repeating 12 samples across different sessions and performing ANOVA. In 95% of cases no significant difference was found across sessions ( p  > 0.05), indicating good panel consistency (Supplementary Table  S2 ).

Aroma and taste perception reported by the trained panel are often linked (Fig.  1 , bottom left panel and Supplementary Data  4 and 5 ), with high correlations between hops aroma and taste (Spearman’s rho=0.83). Bitter taste was found to correlate with hop aroma and taste in general (Spearman’s rho=0.80 and 0.69), and particularly with “grassy” noble hops (Spearman’s rho=0.75). Barnyard flavor, most often associated with sour beers, is identified together with stale hops (Spearman’s rho=0.97) that are used in these beers. Lactic and acetic acid, which often co-occur, are correlated (Spearman’s rho=0.66). Interestingly, sweetness and bitterness are anti-correlated (Spearman’s rho = −0.48), confirming the hypothesis that they mask each other 59 , 60 . Beer body is highly correlated with alcohol (Spearman’s rho = 0.79), and overall appreciation is found to correlate with multiple aspects that describe beer mouthfeel (alcohol, carbonation; Spearman’s rho= 0.32, 0.39), as well as with hop and ester aroma intensity (Spearman’s rho=0.39 and 0.35).

Similar to the chemical analyses, sensorial analyses confirmed typical features of specific beer styles (Supplementary Fig.  S4 ). For example, sour beers (Faro, Flanders Old Brown, Fruit beer, Kriek, Lambic, West Flanders ale) were rated acidic, with flavors of both acetic and lactic acid. Hoppy beers were found to be bitter and showed hop-associated aromas like citrus and tropical fruit. Malt taste is most detected among scotch, stout/porters, and strong ales, while low/no-alcohol beers, which often have a reputation for being ‘worty’ (reminiscent of unfermented, sweet malt extract) appear in the middle. Unsurprisingly, hop aromas are most strongly detected among hoppy beers. Like its chemical counterpart (Supplementary Fig.  S3 ), acidity shows a right-skewed distribution, with the most acidic beers being Krieks, Lambics, and West Flanders ales.

Tasting panel assessments of specific flavors correlate with chemical composition

We find that the concentrations of several chemical compounds strongly correlate with specific aroma or taste, as evaluated by the tasting panel (Fig.  2 , Supplementary Fig.  S5 , Supplementary Data  6 ). In some cases, these correlations confirm expectations and serve as a useful control for data quality. For example, iso-alpha acids, the bittering compounds in hops, strongly correlate with bitterness (Spearman’s rho=0.68), while ethanol and glycerol correlate with tasters’ perceptions of alcohol and body, the mouthfeel sensation of fullness (Spearman’s rho=0.82/0.62 and 0.72/0.57 respectively) and darker color from roasted malts is a good indication of malt perception (Spearman’s rho=0.54).

Heatmap colors indicate Spearman’s Rho. Axes are organized according to sensory categories (aroma, taste, mouthfeel, overall), chemical categories and chemical sources in beer (malt (blue), hops (green), yeast (red), wild flora (yellow), Others (black)). See Supplementary Data  6 for all correlation values.

Interestingly, for some relationships between chemical compounds and perceived flavor, correlations are weaker than expected. For example, the rose-smelling phenethyl acetate only weakly correlates with floral aroma. This hints at more complex relationships and interactions between compounds and suggests a need for a more complex model than simple correlations. Lastly, we uncovered unexpected correlations. For instance, the esters ethyl decanoate and ethyl octanoate appear to correlate slightly with hop perception and bitterness, possibly due to their fruity flavor. Iron is anti-correlated with hop aromas and bitterness, most likely because it is also anti-correlated with iso-alpha acids. This could be a sign of metal chelation of hop acids 61 , given that our analyses measure unbound hop acids and total iron content, or could result from the higher iron content in dark and Fruit beers, which typically have less hoppy and bitter flavors 62 .

Public consumer reviews complement expert panel data

To complement and expand the sensory data of our trained tasting panel, we collected 180,000 reviews of our 250 beers from the online consumer review platform RateBeer. This provided numerical scores for beer appearance, aroma, taste, palate, overall quality as well as the average overall score.

Public datasets are known to suffer from biases, such as price, cult status and psychological conformity towards previous ratings of a product. For example, prices correlate with appreciation scores for these online consumer reviews (rho=0.49, Supplementary Fig.  S6 ), but not for our trained tasting panel (rho=0.19). This suggests that prices affect consumer appreciation, which has been reported in wine 63 , while blind tastings are unaffected. Moreover, we observe that some beer styles, like lagers and non-alcoholic beers, generally receive lower scores, reflecting that online reviewers are mostly beer aficionados with a preference for specialty beers over lager beers. In general, we find a modest correlation between our trained panel’s overall appreciation score and the online consumer appreciation scores (Fig.  3 , rho=0.29). Apart from the aforementioned biases in the online datasets, serving temperature, sample freshness and surroundings, which are all tightly controlled during the tasting panel sessions, can vary tremendously across online consumers and can further contribute to (among others, appreciation) differences between the two categories of tasters. Importantly, in contrast to the overall appreciation scores, for many sensory aspects the results from the professional panel correlated well with results obtained from RateBeer reviews. Correlations were highest for features that are relatively easy to recognize even for untrained tasters, like bitterness, sweetness, alcohol and malt aroma (Fig.  3 and below).

RateBeer text mining results can be found in Supplementary Data  7 . Rho values shown are Spearman correlation values, with asterisks indicating significant correlations ( p  < 0.05, two-sided). All p values were smaller than 0.001, except for Esters aroma (0.0553), Esters taste (0.3275), Esters aroma—banana (0.0019), Coriander (0.0508) and Diacetyl (0.0134).

Besides collecting consumer appreciation from these online reviews, we developed automated text analysis tools to gather additional data from review texts (Supplementary Data  7 ). Processing review texts on the RateBeer database yielded comparable results to the scores given by the trained panel for many common sensory aspects, including acidity, bitterness, sweetness, alcohol, malt, and hop tastes (Fig.  3 ). This is in line with what would be expected, since these attributes require less training for accurate assessment and are less influenced by environmental factors such as temperature, serving glass and odors in the environment. Consumer reviews also correlate well with our trained panel for 4-vinyl guaiacol, a compound associated with a very characteristic aroma. By contrast, correlations for more specific aromas like ester, coriander or diacetyl are underrepresented in the online reviews, underscoring the importance of using a trained tasting panel and standardized tasting sheets with explicit factors to be scored for evaluating specific aspects of a beer. Taken together, our results suggest that public reviews are trustworthy for some, but not all, flavor features and can complement or substitute taste panel data for these sensory aspects.

Models can predict beer sensory profiles from chemical data

The rich datasets of chemical analyses, tasting panel assessments and public reviews gathered in the first part of this study provided us with a unique opportunity to develop predictive models that link chemical data to sensorial features. Given the complexity of beer flavor, basic statistical tools such as correlations or linear regression may not always be the most suitable for making accurate predictions. Instead, we applied different machine learning models that can model both simple linear and complex interactive relationships. Specifically, we constructed a set of regression models to predict (a) trained panel scores for beer flavor and quality and (b) public reviews’ appreciation scores from beer chemical profiles. We trained and tested 10 different models (Methods), 3 linear regression-based models (simple linear regression with first-order interactions (LR), lasso regression with first-order interactions (Lasso), partial least squares regressor (PLSR)), 5 decision tree models (AdaBoost regressor (ABR), extra trees (ET), gradient boosting regressor (GBR), random forest (RF) and XGBoost regressor (XGBR)), 1 support vector regression (SVR), and 1 artificial neural network (ANN) model.

To compare the performance of our machine learning models, the dataset was randomly split into a training and test set, stratified by beer style. After a model was trained on data in the training set, its performance was evaluated on its ability to predict the test dataset obtained from multi-output models (based on the coefficient of determination, see Methods). Additionally, individual-attribute models were ranked per descriptor and the average rank was calculated, as proposed by Korneva et al. 64 . Importantly, both ways of evaluating the models’ performance agreed in general. Performance of the different models varied (Table  1 ). It should be noted that all models perform better at predicting RateBeer results than results from our trained tasting panel. One reason could be that sensory data is inherently variable, and this variability is averaged out with the large number of public reviews from RateBeer. Additionally, all tree-based models perform better at predicting taste than aroma. Linear models (LR) performed particularly poorly, with negative R 2 values, due to severe overfitting (training set R 2  = 1). Overfitting is a common issue in linear models with many parameters and limited samples, especially with interaction terms further amplifying the number of parameters. L1 regularization (Lasso) successfully overcomes this overfitting, out-competing multiple tree-based models on the RateBeer dataset. Similarly, the dimensionality reduction of PLSR avoids overfitting and improves performance, to some extent. Still, tree-based models (ABR, ET, GBR, RF and XGBR) show the best performance, out-competing the linear models (LR, Lasso, PLSR) commonly used in sensory science 65 .

GBR models showed the best overall performance in predicting sensory responses from chemical information, with R 2 values up to 0.75 depending on the predicted sensory feature (Supplementary Table  S4 ). The GBR models predict consumer appreciation (RateBeer) better than our trained panel’s appreciation (R 2 value of 0.67 compared to R 2 value of 0.09) (Supplementary Table  S3 and Supplementary Table  S4 ). ANN models showed intermediate performance, likely because neural networks typically perform best with larger datasets 66 . The SVR shows intermediate performance, mostly due to the weak predictions of specific attributes that lower the overall performance (Supplementary Table  S4 ).

Model dissection identifies specific, unexpected compounds as drivers of consumer appreciation

Next, we leveraged our models to infer important contributors to sensory perception and consumer appreciation. Consumer preference is a crucial sensory aspects, because a product that shows low consumer appreciation scores often does not succeed commercially 25 . Additionally, the requirement for a large number of representative evaluators makes consumer trials one of the more costly and time-consuming aspects of product development. Hence, a model for predicting chemical drivers of overall appreciation would be a welcome addition to the available toolbox for food development and optimization.

Since GBR models on our RateBeer dataset showed the best overall performance, we focused on these models. Specifically, we used two approaches to identify important contributors. First, rankings of the most important predictors for each sensorial trait in the GBR models were obtained based on impurity-based feature importance (mean decrease in impurity). High-ranked parameters were hypothesized to be either the true causal chemical properties underlying the trait, to correlate with the actual causal properties, or to take part in sensory interactions affecting the trait 67 (Fig.  4A ). In a second approach, we used SHAP 68 to determine which parameters contributed most to the model for making predictions of consumer appreciation (Fig.  4B ). SHAP calculates parameter contributions to model predictions on a per-sample basis, which can be aggregated into an importance score.

A The impurity-based feature importance (mean deviance in impurity, MDI) calculated from the Gradient Boosting Regression (GBR) model predicting RateBeer appreciation scores. The top 15 highest ranked chemical properties are shown. B SHAP summary plot for the top 15 parameters contributing to our GBR model. Each point on the graph represents a sample from our dataset. The color represents the concentration of that parameter, with bluer colors representing low values and redder colors representing higher values. Greater absolute values on the horizontal axis indicate a higher impact of the parameter on the prediction of the model. C Spearman correlations between the 15 most important chemical properties and consumer overall appreciation. Numbers indicate the Spearman Rho correlation coefficient, and the rank of this correlation compared to all other correlations. The top 15 important compounds were determined using SHAP (panel B).

Both approaches identified ethyl acetate as the most predictive parameter for beer appreciation (Fig.  4 ). Ethyl acetate is the most abundant ester in beer with a typical ‘fruity’, ‘solvent’ and ‘alcoholic’ flavor, but is often considered less important than other esters like isoamyl acetate. The second most important parameter identified by SHAP is ethanol, the most abundant beer compound after water. Apart from directly contributing to beer flavor and mouthfeel, ethanol drastically influences the physical properties of beer, dictating how easily volatile compounds escape the beer matrix to contribute to beer aroma 69 . Importantly, it should also be noted that the importance of ethanol for appreciation is likely inflated by the very low appreciation scores of non-alcoholic beers (Supplementary Fig.  S4 ). Despite not often being considered a driver of beer appreciation, protein level also ranks highly in both approaches, possibly due to its effect on mouthfeel and body 70 . Lactic acid, which contributes to the tart taste of sour beers, is the fourth most important parameter identified by SHAP, possibly due to the generally high appreciation of sour beers in our dataset.

Interestingly, some of the most important predictive parameters for our model are not well-established as beer flavors or are even commonly regarded as being negative for beer quality. For example, our models identify methanethiol and ethyl phenyl acetate, an ester commonly linked to beer staling 71 , as a key factor contributing to beer appreciation. Although there is no doubt that high concentrations of these compounds are considered unpleasant, the positive effects of modest concentrations are not yet known 72 , 73 .

To compare our approach to conventional statistics, we evaluated how well the 15 most important SHAP-derived parameters correlate with consumer appreciation (Fig.  4C ). Interestingly, only 6 of the properties derived by SHAP rank amongst the top 15 most correlated parameters. For some chemical compounds, the correlations are so low that they would have likely been considered unimportant. For example, lactic acid, the fourth most important parameter, shows a bimodal distribution for appreciation, with sour beers forming a separate cluster, that is missed entirely by the Spearman correlation. Additionally, the correlation plots reveal outliers, emphasizing the need for robust analysis tools. Together, this highlights the need for alternative models, like the Gradient Boosting model, that better grasp the complexity of (beer) flavor.

Finally, to observe the relationships between these chemical properties and their predicted targets, partial dependence plots were constructed for the six most important predictors of consumer appreciation 74 , 75 , 76 (Supplementary Fig.  S7 ). One-way partial dependence plots show how a change in concentration affects the predicted appreciation. These plots reveal an important limitation of our models: appreciation predictions remain constant at ever-increasing concentrations. This implies that once a threshold concentration is reached, further increasing the concentration does not affect appreciation. This is false, as it is well-documented that certain compounds become unpleasant at high concentrations, including ethyl acetate (‘nail polish’) 77 and methanethiol (‘sulfury’ and ‘rotten cabbage’) 78 . The inability of our models to grasp that flavor compounds have optimal levels, above which they become negative, is a consequence of working with commercial beer brands where (off-)flavors are rarely too high to negatively impact the product. The two-way partial dependence plots show how changing the concentration of two compounds influences predicted appreciation, visualizing their interactions (Supplementary Fig.  S7 ). In our case, the top 5 parameters are dominated by additive or synergistic interactions, with high concentrations for both compounds resulting in the highest predicted appreciation.

To assess the robustness of our best-performing models and model predictions, we performed 100 iterations of the GBR, RF and ET models. In general, all iterations of the models yielded similar performance (Supplementary Fig.  S8 ). Moreover, the main predictors (including the top predictors ethanol and ethyl acetate) remained virtually the same, especially for GBR and RF. For the iterations of the ET model, we did observe more variation in the top predictors, which is likely a consequence of the model’s inherent random architecture in combination with co-correlations between certain predictors. However, even in this case, several of the top predictors (ethanol and ethyl acetate) remain unchanged, although their rank in importance changes (Supplementary Fig.  S8 ).

Next, we investigated if a combination of RateBeer and trained panel data into one consolidated dataset would lead to stronger models, under the hypothesis that such a model would suffer less from bias in the datasets. A GBR model was trained to predict appreciation on the combined dataset. This model underperformed compared to the RateBeer model, both in the native case and when including a dataset identifier (R 2  = 0.67, 0.26 and 0.42 respectively). For the latter, the dataset identifier is the most important feature (Supplementary Fig.  S9 ), while most of the feature importance remains unchanged, with ethyl acetate and ethanol ranking highest, like in the original model trained only on RateBeer data. It seems that the large variation in the panel dataset introduces noise, weakening the models’ performances and reliability. In addition, it seems reasonable to assume that both datasets are fundamentally different, with the panel dataset obtained by blind tastings by a trained professional panel.

Lastly, we evaluated whether beer style identifiers would further enhance the model’s performance. A GBR model was trained with parameters that explicitly encoded the styles of the samples. This did not improve model performance (R2 = 0.66 with style information vs R2 = 0.67). The most important chemical features are consistent with the model trained without style information (eg. ethanol and ethyl acetate), and with the exception of the most preferred (strong ale) and least preferred (low/no-alcohol) styles, none of the styles were among the most important features (Supplementary Fig.  S9 , Supplementary Table  S5 and S6 ). This is likely due to a combination of style-specific chemical signatures, such as iso-alpha acids and lactic acid, that implicitly convey style information to the original models, as well as the low number of samples belonging to some styles, making it difficult for the model to learn style-specific patterns. Moreover, beer styles are not rigorously defined, with some styles overlapping in features and some beers being misattributed to a specific style, all of which leads to more noise in models that use style parameters.

Model validation

To test if our predictive models give insight into beer appreciation, we set up experiments aimed at improving existing commercial beers. We specifically selected overall appreciation as the trait to be examined because of its complexity and commercial relevance. Beer flavor comprises a complex bouquet rather than single aromas and tastes 53 . Hence, adding a single compound to the extent that a difference is noticeable may lead to an unbalanced, artificial flavor. Therefore, we evaluated the effect of combinations of compounds. Because Blond beers represent the most extensive style in our dataset, we selected a beer from this style as the starting material for these experiments (Beer 64 in Supplementary Data  1 ).

In the first set of experiments, we adjusted the concentrations of compounds that made up the most important predictors of overall appreciation (ethyl acetate, ethanol, lactic acid, ethyl phenyl acetate) together with correlated compounds (ethyl hexanoate, isoamyl acetate, glycerol), bringing them up to 95 th percentile ethanol-normalized concentrations (Methods) within the Blond group (‘Spiked’ concentration in Fig.  5A ). Compared to controls, the spiked beers were found to have significantly improved overall appreciation among trained panelists, with panelist noting increased intensity of ester flavors, sweetness, alcohol, and body fullness (Fig.  5B ). To disentangle the contribution of ethanol to these results, a second experiment was performed without the addition of ethanol. This resulted in a similar outcome, including increased perception of alcohol and overall appreciation.

Adding the top chemical compounds, identified as best predictors of appreciation by our model, into poorly appreciated beers results in increased appreciation from our trained panel. Results of sensory tests between base beers and those spiked with compounds identified as the best predictors by the model. A Blond and Non/Low-alcohol (0.0% ABV) base beers were brought up to 95th-percentile ethanol-normalized concentrations within each style. B For each sensory attribute, tasters indicated the more intense sample and selected the sample they preferred. The numbers above the bars correspond to the p values that indicate significant changes in perceived flavor (two-sided binomial test: alpha 0.05, n  = 20 or 13).

In a last experiment, we tested whether using the model’s predictions can boost the appreciation of a non-alcoholic beer (beer 223 in Supplementary Data  1 ). Again, the addition of a mixture of predicted compounds (omitting ethanol, in this case) resulted in a significant increase in appreciation, body, ester flavor and sweetness.

Predicting flavor and consumer appreciation from chemical composition is one of the ultimate goals of sensory science. A reliable, systematic and unbiased way to link chemical profiles to flavor and food appreciation would be a significant asset to the food and beverage industry. Such tools would substantially aid in quality control and recipe development, offer an efficient and cost-effective alternative to pilot studies and consumer trials and would ultimately allow food manufacturers to produce superior, tailor-made products that better meet the demands of specific consumer groups more efficiently.

A limited set of studies have previously tried, to varying degrees of success, to predict beer flavor and beer popularity based on (a limited set of) chemical compounds and flavors 79 , 80 . Current sensitive, high-throughput technologies allow measuring an unprecedented number of chemical compounds and properties in a large set of samples, yielding a dataset that can train models that help close the gaps between chemistry and flavor, even for a complex natural product like beer. To our knowledge, no previous research gathered data at this scale (250 samples, 226 chemical parameters, 50 sensory attributes and 5 consumer scores) to disentangle and validate the chemical aspects driving beer preference using various machine-learning techniques. We find that modern machine learning models outperform conventional statistical tools, such as correlations and linear models, and can successfully predict flavor appreciation from chemical composition. This could be attributed to the natural incorporation of interactions and non-linear or discontinuous effects in machine learning models, which are not easily grasped by the linear model architecture. While linear models and partial least squares regression represent the most widespread statistical approaches in sensory science, in part because they allow interpretation 65 , 81 , 82 , modern machine learning methods allow for building better predictive models while preserving the possibility to dissect and exploit the underlying patterns. Of the 10 different models we trained, tree-based models, such as our best performing GBR, showed the best overall performance in predicting sensory responses from chemical information, outcompeting artificial neural networks. This agrees with previous reports for models trained on tabular data 83 . Our results are in line with the findings of Colantonio et al. who also identified the gradient boosting architecture as performing best at predicting appreciation and flavor (of tomatoes and blueberries, in their specific study) 26 . Importantly, besides our larger experimental scale, we were able to directly confirm our models’ predictions in vivo.

Our study confirms that flavor compound concentration does not always correlate with perception, suggesting complex interactions that are often missed by more conventional statistics and simple models. Specifically, we find that tree-based algorithms may perform best in developing models that link complex food chemistry with aroma. Furthermore, we show that massive datasets of untrained consumer reviews provide a valuable source of data, that can complement or even replace trained tasting panels, especially for appreciation and basic flavors, such as sweetness and bitterness. This holds despite biases that are known to occur in such datasets, such as price or conformity bias. Moreover, GBR models predict taste better than aroma. This is likely because taste (e.g. bitterness) often directly relates to the corresponding chemical measurements (e.g., iso-alpha acids), whereas such a link is less clear for aromas, which often result from the interplay between multiple volatile compounds. We also find that our models are best at predicting acidity and alcohol, likely because there is a direct relation between the measured chemical compounds (acids and ethanol) and the corresponding perceived sensorial attribute (acidity and alcohol), and because even untrained consumers are generally able to recognize these flavors and aromas.

The predictions of our final models, trained on review data, hold even for blind tastings with small groups of trained tasters, as demonstrated by our ability to validate specific compounds as drivers of beer flavor and appreciation. Since adding a single compound to the extent of a noticeable difference may result in an unbalanced flavor profile, we specifically tested our identified key drivers as a combination of compounds. While this approach does not allow us to validate if a particular single compound would affect flavor and/or appreciation, our experiments do show that this combination of compounds increases consumer appreciation.

It is important to stress that, while it represents an important step forward, our approach still has several major limitations. A key weakness of the GBR model architecture is that amongst co-correlating variables, the largest main effect is consistently preferred for model building. As a result, co-correlating variables often have artificially low importance scores, both for impurity and SHAP-based methods, like we observed in the comparison to the more randomized Extra Trees models. This implies that chemicals identified as key drivers of a specific sensory feature by GBR might not be the true causative compounds, but rather co-correlate with the actual causative chemical. For example, the high importance of ethyl acetate could be (partially) attributed to the total ester content, ethanol or ethyl hexanoate (rho=0.77, rho=0.72 and rho=0.68), while ethyl phenylacetate could hide the importance of prenyl isobutyrate and ethyl benzoate (rho=0.77 and rho=0.76). Expanding our GBR model to include beer style as a parameter did not yield additional power or insight. This is likely due to style-specific chemical signatures, such as iso-alpha acids and lactic acid, that implicitly convey style information to the original model, as well as the smaller sample size per style, limiting the power to uncover style-specific patterns. This can be partly attributed to the curse of dimensionality, where the high number of parameters results in the models mainly incorporating single parameter effects, rather than complex interactions such as style-dependent effects 67 . A larger number of samples may overcome some of these limitations and offer more insight into style-specific effects. On the other hand, beer style is not a rigid scientific classification, and beers within one style often differ a lot, which further complicates the analysis of style as a model factor.

Our study is limited to beers from Belgian breweries. Although these beers cover a large portion of the beer styles available globally, some beer styles and consumer patterns may be missing, while other features might be overrepresented. For example, many Belgian ales exhibit yeast-driven flavor profiles, which is reflected in the chemical drivers of appreciation discovered by this study. In future work, expanding the scope to include diverse markets and beer styles could lead to the identification of even more drivers of appreciation and better models for special niche products that were not present in our beer set.

In addition to inherent limitations of GBR models, there are also some limitations associated with studying food aroma. Even if our chemical analyses measured most of the known aroma compounds, the total number of flavor compounds in complex foods like beer is still larger than the subset we were able to measure in this study. For example, hop-derived thiols, that influence flavor at very low concentrations, are notoriously difficult to measure in a high-throughput experiment. Moreover, consumer perception remains subjective and prone to biases that are difficult to avoid. It is also important to stress that the models are still immature and that more extensive datasets will be crucial for developing more complete models in the future. Besides more samples and parameters, our dataset does not include any demographic information about the tasters. Including such data could lead to better models that grasp external factors like age and culture. Another limitation is that our set of beers consists of high-quality end-products and lacks beers that are unfit for sale, which limits the current model in accurately predicting products that are appreciated very badly. Finally, while models could be readily applied in quality control, their use in sensory science and product development is restrained by their inability to discern causal relationships. Given that the models cannot distinguish compounds that genuinely drive consumer perception from those that merely correlate, validation experiments are essential to identify true causative compounds.

Despite the inherent limitations, dissection of our models enabled us to pinpoint specific molecules as potential drivers of beer aroma and consumer appreciation, including compounds that were unexpected and would not have been identified using standard approaches. Important drivers of beer appreciation uncovered by our models include protein levels, ethyl acetate, ethyl phenyl acetate and lactic acid. Currently, many brewers already use lactic acid to acidify their brewing water and ensure optimal pH for enzymatic activity during the mashing process. Our results suggest that adding lactic acid can also improve beer appreciation, although its individual effect remains to be tested. Interestingly, ethanol appears to be unnecessary to improve beer appreciation, both for blond beer and alcohol-free beer. Given the growing consumer interest in alcohol-free beer, with a predicted annual market growth of >7% 84 , it is relevant for brewers to know what compounds can further increase consumer appreciation of these beers. Hence, our model may readily provide avenues to further improve the flavor and consumer appreciation of both alcoholic and non-alcoholic beers, which is generally considered one of the key challenges for future beer production.

Whereas we see a direct implementation of our results for the development of superior alcohol-free beverages and other food products, our study can also serve as a stepping stone for the development of novel alcohol-containing beverages. We want to echo the growing body of scientific evidence for the negative effects of alcohol consumption, both on the individual level by the mutagenic, teratogenic and carcinogenic effects of ethanol 85 , 86 , as well as the burden on society caused by alcohol abuse and addiction. We encourage the use of our results for the production of healthier, tastier products, including novel and improved beverages with lower alcohol contents. Furthermore, we strongly discourage the use of these technologies to improve the appreciation or addictive properties of harmful substances.

The present work demonstrates that despite some important remaining hurdles, combining the latest developments in chemical analyses, sensory analysis and modern machine learning methods offers exciting avenues for food chemistry and engineering. Soon, these tools may provide solutions in quality control and recipe development, as well as new approaches to sensory science and flavor research.

Beer selection

250 commercial Belgian beers were selected to cover the broad diversity of beer styles and corresponding diversity in chemical composition and aroma. See Supplementary Fig.  S1 .

Chemical dataset

Sample preparation.

Beers within their expiration date were purchased from commercial retailers. Samples were prepared in biological duplicates at room temperature, unless explicitly stated otherwise. Bottle pressure was measured with a manual pressure device (Steinfurth Mess-Systeme GmbH) and used to calculate CO 2 concentration. The beer was poured through two filter papers (Macherey-Nagel, 500713032 MN 713 ¼) to remove carbon dioxide and prevent spontaneous foaming. Samples were then prepared for measurements by targeted Headspace-Gas Chromatography-Flame Ionization Detector/Flame Photometric Detector (HS-GC-FID/FPD), Headspace-Solid Phase Microextraction-Gas Chromatography-Mass Spectrometry (HS-SPME-GC-MS), colorimetric analysis, enzymatic analysis, Near-Infrared (NIR) analysis, as described in the sections below. The mean values of biological duplicates are reported for each compound.

HS-GC-FID/FPD

HS-GC-FID/FPD (Shimadzu GC 2010 Plus) was used to measure higher alcohols, acetaldehyde, esters, 4-vinyl guaicol, and sulfur compounds. Each measurement comprised 5 ml of sample pipetted into a 20 ml glass vial containing 1.75 g NaCl (VWR, 27810.295). 100 µl of 2-heptanol (Sigma-Aldrich, H3003) (internal standard) solution in ethanol (Fisher Chemical, E/0650DF/C17) was added for a final concentration of 2.44 mg/L. Samples were flushed with nitrogen for 10 s, sealed with a silicone septum, stored at −80 °C and analyzed in batches of 20.

The GC was equipped with a DB-WAXetr column (length, 30 m; internal diameter, 0.32 mm; layer thickness, 0.50 µm; Agilent Technologies, Santa Clara, CA, USA) to the FID and an HP-5 column (length, 30 m; internal diameter, 0.25 mm; layer thickness, 0.25 µm; Agilent Technologies, Santa Clara, CA, USA) to the FPD. N 2 was used as the carrier gas. Samples were incubated for 20 min at 70 °C in the headspace autosampler (Flow rate, 35 cm/s; Injection volume, 1000 µL; Injection mode, split; Combi PAL autosampler, CTC analytics, Switzerland). The injector, FID and FPD temperatures were kept at 250 °C. The GC oven temperature was first held at 50 °C for 5 min and then allowed to rise to 80 °C at a rate of 5 °C/min, followed by a second ramp of 4 °C/min until 200 °C kept for 3 min and a final ramp of (4 °C/min) until 230 °C for 1 min. Results were analyzed with the GCSolution software version 2.4 (Shimadzu, Kyoto, Japan). The GC was calibrated with a 5% EtOH solution (VWR International) containing the volatiles under study (Supplementary Table  S7 ).

HS-SPME-GC-MS

HS-SPME-GC-MS (Shimadzu GCMS-QP-2010 Ultra) was used to measure additional volatile compounds, mainly comprising terpenoids and esters. Samples were analyzed by HS-SPME using a triphase DVB/Carboxen/PDMS 50/30 μm SPME fiber (Supelco Co., Bellefonte, PA, USA) followed by gas chromatography (Thermo Fisher Scientific Trace 1300 series, USA) coupled to a mass spectrometer (Thermo Fisher Scientific ISQ series MS) equipped with a TriPlus RSH autosampler. 5 ml of degassed beer sample was placed in 20 ml vials containing 1.75 g NaCl (VWR, 27810.295). 5 µl internal standard mix was added, containing 2-heptanol (1 g/L) (Sigma-Aldrich, H3003), 4-fluorobenzaldehyde (1 g/L) (Sigma-Aldrich, 128376), 2,3-hexanedione (1 g/L) (Sigma-Aldrich, 144169) and guaiacol (1 g/L) (Sigma-Aldrich, W253200) in ethanol (Fisher Chemical, E/0650DF/C17). Each sample was incubated at 60 °C in the autosampler oven with constant agitation. After 5 min equilibration, the SPME fiber was exposed to the sample headspace for 30 min. The compounds trapped on the fiber were thermally desorbed in the injection port of the chromatograph by heating the fiber for 15 min at 270 °C.

The GC-MS was equipped with a low polarity RXi-5Sil MS column (length, 20 m; internal diameter, 0.18 mm; layer thickness, 0.18 µm; Restek, Bellefonte, PA, USA). Injection was performed in splitless mode at 320 °C, a split flow of 9 ml/min, a purge flow of 5 ml/min and an open valve time of 3 min. To obtain a pulsed injection, a programmed gas flow was used whereby the helium gas flow was set at 2.7 mL/min for 0.1 min, followed by a decrease in flow of 20 ml/min to the normal 0.9 mL/min. The temperature was first held at 30 °C for 3 min and then allowed to rise to 80 °C at a rate of 7 °C/min, followed by a second ramp of 2 °C/min till 125 °C and a final ramp of 8 °C/min with a final temperature of 270 °C.

Mass acquisition range was 33 to 550 amu at a scan rate of 5 scans/s. Electron impact ionization energy was 70 eV. The interface and ion source were kept at 275 °C and 250 °C, respectively. A mix of linear n-alkanes (from C7 to C40, Supelco Co.) was injected into the GC-MS under identical conditions to serve as external retention index markers. Identification and quantification of the compounds were performed using an in-house developed R script as described in Goelen et al. and Reher et al. 87 , 88 (for package information, see Supplementary Table  S8 ). Briefly, chromatograms were analyzed using AMDIS (v2.71) 89 to separate overlapping peaks and obtain pure compound spectra. The NIST MS Search software (v2.0 g) in combination with the NIST2017, FFNSC3 and Adams4 libraries were used to manually identify the empirical spectra, taking into account the expected retention time. After background subtraction and correcting for retention time shifts between samples run on different days based on alkane ladders, compound elution profiles were extracted and integrated using a file with 284 target compounds of interest, which were either recovered in our identified AMDIS list of spectra or were known to occur in beer. Compound elution profiles were estimated for every peak in every chromatogram over a time-restricted window using weighted non-negative least square analysis after which peak areas were integrated 87 , 88 . Batch effect correction was performed by normalizing against the most stable internal standard compound, 4-fluorobenzaldehyde. Out of all 284 target compounds that were analyzed, 167 were visually judged to have reliable elution profiles and were used for final analysis.

Discrete photometric and enzymatic analysis

Discrete photometric and enzymatic analysis (Thermo Scientific TM Gallery TM Plus Beermaster Discrete Analyzer) was used to measure acetic acid, ammonia, beta-glucan, iso-alpha acids, color, sugars, glycerol, iron, pH, protein, and sulfite. 2 ml of sample volume was used for the analyses. Information regarding the reagents and standard solutions used for analyses and calibrations is included in Supplementary Table  S7 and Supplementary Table  S9 .

NIR analyses

NIR analysis (Anton Paar Alcolyzer Beer ME System) was used to measure ethanol. Measurements comprised 50 ml of sample, and a 10% EtOH solution was used for calibration.

Correlation calculations

Pairwise Spearman Rank correlations were calculated between all chemical properties.

Sensory dataset

Trained panel.

Our trained tasting panel consisted of volunteers who gave prior verbal informed consent. All compounds used for the validation experiment were of food-grade quality. The tasting sessions were approved by the Social and Societal Ethics Committee of the KU Leuven (G-2022-5677-R2(MAR)). All online reviewers agreed to the Terms and Conditions of the RateBeer website.

Sensory analysis was performed according to the American Society of Brewing Chemists (ASBC) Sensory Analysis Methods 90 . 30 volunteers were screened through a series of triangle tests. The sixteen most sensitive and consistent tasters were retained as taste panel members. The resulting panel was diverse in age [22–42, mean: 29], sex [56% male] and nationality [7 different countries]. The panel developed a consensus vocabulary to describe beer aroma, taste and mouthfeel. Panelists were trained to identify and score 50 different attributes, using a 7-point scale to rate attributes’ intensity. The scoring sheet is included as Supplementary Data  3 . Sensory assessments took place between 10–12 a.m. The beers were served in black-colored glasses. Per session, between 5 and 12 beers of the same style were tasted at 12 °C to 16 °C. Two reference beers were added to each set and indicated as ‘Reference 1 & 2’, allowing panel members to calibrate their ratings. Not all panelists were present at every tasting. Scores were scaled by standard deviation and mean-centered per taster. Values are represented as z-scores and clustered by Euclidean distance. Pairwise Spearman correlations were calculated between taste and aroma sensory attributes. Panel consistency was evaluated by repeating samples on different sessions and performing ANOVA to identify differences, using the ‘stats’ package (v4.2.2) in R (for package information, see Supplementary Table  S8 ).

Online reviews from a public database

The ‘scrapy’ package in Python (v3.6) (for package information, see Supplementary Table  S8 ). was used to collect 232,288 online reviews (mean=922, min=6, max=5343) from RateBeer, an online beer review database. Each review entry comprised 5 numerical scores (appearance, aroma, taste, palate and overall quality) and an optional review text. The total number of reviews per reviewer was collected separately. Numerical scores were scaled and centered per rater, and mean scores were calculated per beer.

For the review texts, the language was estimated using the packages ‘langdetect’ and ‘langid’ in Python. Reviews that were classified as English by both packages were kept. Reviewers with fewer than 100 entries overall were discarded. 181,025 reviews from >6000 reviewers from >40 countries remained. Text processing was done using the ‘nltk’ package in Python. Texts were corrected for slang and misspellings; proper nouns and rare words that are relevant to the beer context were specified and kept as-is (‘Chimay’,’Lambic’, etc.). A dictionary of semantically similar sensorial terms, for example ‘floral’ and ‘flower’, was created and collapsed together into one term. Words were stemmed and lemmatized to avoid identifying words such as ‘acid’ and ‘acidity’ as separate terms. Numbers and punctuation were removed.

Sentences from up to 50 randomly chosen reviews per beer were manually categorized according to the aspect of beer they describe (appearance, aroma, taste, palate, overall quality—not to be confused with the 5 numerical scores described above) or flagged as irrelevant if they contained no useful information. If a beer contained fewer than 50 reviews, all reviews were manually classified. This labeled data set was used to train a model that classified the rest of the sentences for all beers 91 . Sentences describing taste and aroma were extracted, and term frequency–inverse document frequency (TFIDF) was implemented to calculate enrichment scores for sensorial words per beer.

The sex of the tasting subject was not considered when building our sensory database. Instead, results from different panelists were averaged, both for our trained panel (56% male, 44% female) and the RateBeer reviews (70% male, 30% female for RateBeer as a whole).

Beer price collection and processing

Beer prices were collected from the following stores: Colruyt, Delhaize, Total Wine, BeerHawk, The Belgian Beer Shop, The Belgian Shop, and Beer of Belgium. Where applicable, prices were converted to Euros and normalized per liter. Spearman correlations were calculated between these prices and mean overall appreciation scores from RateBeer and the taste panel, respectively.

Pairwise Spearman Rank correlations were calculated between all sensory properties.

Machine learning models

Predictive modeling of sensory profiles from chemical data.

Regression models were constructed to predict (a) trained panel scores for beer flavors and quality from beer chemical profiles and (b) public reviews’ appreciation scores from beer chemical profiles. Z-scores were used to represent sensory attributes in both data sets. Chemical properties with log-normal distributions (Shapiro-Wilk test, p  <  0.05 ) were log-transformed. Missing chemical measurements (0.1% of all data) were replaced with mean values per attribute. Observations from 250 beers were randomly separated into a training set (70%, 175 beers) and a test set (30%, 75 beers), stratified per beer style. Chemical measurements (p = 231) were normalized based on the training set average and standard deviation. In total, three linear regression-based models: linear regression with first-order interaction terms (LR), lasso regression with first-order interaction terms (Lasso) and partial least squares regression (PLSR); five decision tree models, Adaboost regressor (ABR), Extra Trees (ET), Gradient Boosting regressor (GBR), Random Forest (RF) and XGBoost regressor (XGBR); one support vector machine model (SVR) and one artificial neural network model (ANN) were trained. The models were implemented using the ‘scikit-learn’ package (v1.2.2) and ‘xgboost’ package (v1.7.3) in Python (v3.9.16). Models were trained, and hyperparameters optimized, using five-fold cross-validated grid search with the coefficient of determination (R 2 ) as the evaluation metric. The ANN (scikit-learn’s MLPRegressor) was optimized using Bayesian Tree-Structured Parzen Estimator optimization with the ‘Optuna’ Python package (v3.2.0). Individual models were trained per attribute, and a multi-output model was trained on all attributes simultaneously.

Model dissection

GBR was found to outperform other methods, resulting in models with the highest average R 2 values in both trained panel and public review data sets. Impurity-based rankings of the most important predictors for each predicted sensorial trait were obtained using the ‘scikit-learn’ package. To observe the relationships between these chemical properties and their predicted targets, partial dependence plots (PDP) were constructed for the six most important predictors of consumer appreciation 74 , 75 .

The ‘SHAP’ package in Python (v0.41.0) was implemented to provide an alternative ranking of predictor importance and to visualize the predictors’ effects as a function of their concentration 68 .

Validation of causal chemical properties

To validate the effects of the most important model features on predicted sensory attributes, beers were spiked with the chemical compounds identified by the models and descriptive sensory analyses were carried out according to the American Society of Brewing Chemists (ASBC) protocol 90 .

Compound spiking was done 30 min before tasting. Compounds were spiked into fresh beer bottles, that were immediately resealed and inverted three times. Fresh bottles of beer were opened for the same duration, resealed, and inverted thrice, to serve as controls. Pairs of spiked samples and controls were served simultaneously, chilled and in dark glasses as outlined in the Trained panel section above. Tasters were instructed to select the glass with the higher flavor intensity for each attribute (directional difference test 92 ) and to select the glass they prefer.

The final concentration after spiking was equal to the within-style average, after normalizing by ethanol concentration. This was done to ensure balanced flavor profiles in the final spiked beer. The same methods were applied to improve a non-alcoholic beer. Compounds were the following: ethyl acetate (Merck KGaA, W241415), ethyl hexanoate (Merck KGaA, W243906), isoamyl acetate (Merck KGaA, W205508), phenethyl acetate (Merck KGaA, W285706), ethanol (96%, Colruyt), glycerol (Merck KGaA, W252506), lactic acid (Merck KGaA, 261106).

Significant differences in preference or perceived intensity were determined by performing the two-sided binomial test on each attribute.

Reporting summary

Further information on research design is available in the  Nature Portfolio Reporting Summary linked to this article.

Data availability

The data that support the findings of this work are available in the Supplementary Data files and have been deposited to Zenodo under accession code 10653704 93 . The RateBeer scores data are under restricted access, they are not publicly available as they are property of RateBeer (ZX Ventures, USA). Access can be obtained from the authors upon reasonable request and with permission of RateBeer (ZX Ventures, USA).  Source data are provided with this paper.

Code availability

The code for training the machine learning models, analyzing the models, and generating the figures has been deposited to Zenodo under accession code 10653704 93 .

Tieman, D. et al. A chemical genetic roadmap to improved tomato flavor. Science 355 , 391–394 (2017).

Article   ADS   CAS   PubMed   Google Scholar  

Plutowska, B. & Wardencki, W. Application of gas chromatography–olfactometry (GC–O) in analysis and quality assessment of alcoholic beverages – A review. Food Chem. 107 , 449–463 (2008).

Article   CAS   Google Scholar  

Legin, A., Rudnitskaya, A., Seleznev, B. & Vlasov, Y. Electronic tongue for quality assessment of ethanol, vodka and eau-de-vie. Anal. Chim. Acta 534 , 129–135 (2005).

Loutfi, A., Coradeschi, S., Mani, G. K., Shankar, P. & Rayappan, J. B. B. Electronic noses for food quality: A review. J. Food Eng. 144 , 103–111 (2015).

Ahn, Y.-Y., Ahnert, S. E., Bagrow, J. P. & Barabási, A.-L. Flavor network and the principles of food pairing. Sci. Rep. 1 , 196 (2011).

Article   CAS   PubMed   PubMed Central   Google Scholar  

Bartoshuk, L. M. & Klee, H. J. Better fruits and vegetables through sensory analysis. Curr. Biol. 23 , R374–R378 (2013).

Article   CAS   PubMed   Google Scholar  

Piggott, J. R. Design questions in sensory and consumer science. Food Qual. Prefer. 3293 , 217–220 (1995).

Article   Google Scholar  

Kermit, M. & Lengard, V. Assessing the performance of a sensory panel-panellist monitoring and tracking. J. Chemom. 19 , 154–161 (2005).

Cook, D. J., Hollowood, T. A., Linforth, R. S. T. & Taylor, A. J. Correlating instrumental measurements of texture and flavour release with human perception. Int. J. Food Sci. Technol. 40 , 631–641 (2005).

Chinchanachokchai, S., Thontirawong, P. & Chinchanachokchai, P. A tale of two recommender systems: The moderating role of consumer expertise on artificial intelligence based product recommendations. J. Retail. Consum. Serv. 61 , 1–12 (2021).

Ross, C. F. Sensory science at the human-machine interface. Trends Food Sci. Technol. 20 , 63–72 (2009).

Chambers, E. IV & Koppel, K. Associations of volatile compounds with sensory aroma and flavor: The complex nature of flavor. Molecules 18 , 4887–4905 (2013).

Pinu, F. R. Metabolomics—The new frontier in food safety and quality research. Food Res. Int. 72 , 80–81 (2015).

Danezis, G. P., Tsagkaris, A. S., Brusic, V. & Georgiou, C. A. Food authentication: state of the art and prospects. Curr. Opin. Food Sci. 10 , 22–31 (2016).

Shepherd, G. M. Smell images and the flavour system in the human brain. Nature 444 , 316–321 (2006).

Meilgaard, M. C. Prediction of flavor differences between beers from their chemical composition. J. Agric. Food Chem. 30 , 1009–1017 (1982).

Xu, L. et al. Widespread receptor-driven modulation in peripheral olfactory coding. Science 368 , eaaz5390 (2020).

Kupferschmidt, K. Following the flavor. Science 340 , 808–809 (2013).

Billesbølle, C. B. et al. Structural basis of odorant recognition by a human odorant receptor. Nature 615 , 742–749 (2023).

Article   ADS   PubMed   PubMed Central   Google Scholar  

Smith, B. Perspective: Complexities of flavour. Nature 486 , S6–S6 (2012).

Pfister, P. et al. Odorant receptor inhibition is fundamental to odor encoding. Curr. Biol. 30 , 2574–2587 (2020).

Moskowitz, H. W., Kumaraiah, V., Sharma, K. N., Jacobs, H. L. & Sharma, S. D. Cross-cultural differences in simple taste preferences. Science 190 , 1217–1218 (1975).

Eriksson, N. et al. A genetic variant near olfactory receptor genes influences cilantro preference. Flavour 1 , 22 (2012).

Ferdenzi, C. et al. Variability of affective responses to odors: Culture, gender, and olfactory knowledge. Chem. Senses 38 , 175–186 (2013).

Article   PubMed   Google Scholar  

Lawless, H. T. & Heymann, H. Sensory evaluation of food: Principles and practices. (Springer, New York, NY). https://doi.org/10.1007/978-1-4419-6488-5 (2010).

Colantonio, V. et al. Metabolomic selection for enhanced fruit flavor. Proc. Natl. Acad. Sci. 119 , e2115865119 (2022).

Fritz, F., Preissner, R. & Banerjee, P. VirtualTaste: a web server for the prediction of organoleptic properties of chemical compounds. Nucleic Acids Res 49 , W679–W684 (2021).

Tuwani, R., Wadhwa, S. & Bagler, G. BitterSweet: Building machine learning models for predicting the bitter and sweet taste of small molecules. Sci. Rep. 9 , 1–13 (2019).

Dagan-Wiener, A. et al. Bitter or not? BitterPredict, a tool for predicting taste from chemical structure. Sci. Rep. 7 , 1–13 (2017).

Pallante, L. et al. Toward a general and interpretable umami taste predictor using a multi-objective machine learning approach. Sci. Rep. 12 , 1–11 (2022).

Malavolta, M. et al. A survey on computational taste predictors. Eur. Food Res. Technol. 248 , 2215–2235 (2022).

Lee, B. K. et al. A principal odor map unifies diverse tasks in olfactory perception. Science 381 , 999–1006 (2023).

Mayhew, E. J. et al. Transport features predict if a molecule is odorous. Proc. Natl. Acad. Sci. 119 , e2116576119 (2022).

Niu, Y. et al. Sensory evaluation of the synergism among ester odorants in light aroma-type liquor by odor threshold, aroma intensity and flash GC electronic nose. Food Res. Int. 113 , 102–114 (2018).

Yu, P., Low, M. Y. & Zhou, W. Design of experiments and regression modelling in food flavour and sensory analysis: A review. Trends Food Sci. Technol. 71 , 202–215 (2018).

Oladokun, O. et al. The impact of hop bitter acid and polyphenol profiles on the perceived bitterness of beer. Food Chem. 205 , 212–220 (2016).

Linforth, R., Cabannes, M., Hewson, L., Yang, N. & Taylor, A. Effect of fat content on flavor delivery during consumption: An in vivo model. J. Agric. Food Chem. 58 , 6905–6911 (2010).

Guo, S., Na Jom, K. & Ge, Y. Influence of roasting condition on flavor profile of sunflower seeds: A flavoromics approach. Sci. Rep. 9 , 11295 (2019).

Ren, Q. et al. The changes of microbial community and flavor compound in the fermentation process of Chinese rice wine using Fagopyrum tataricum grain as feedstock. Sci. Rep. 9 , 3365 (2019).

Hastie, T., Friedman, J. & Tibshirani, R. The Elements of Statistical Learning. (Springer, New York, NY). https://doi.org/10.1007/978-0-387-21606-5 (2001).

Dietz, C., Cook, D., Huismann, M., Wilson, C. & Ford, R. The multisensory perception of hop essential oil: a review. J. Inst. Brew. 126 , 320–342 (2020).

CAS   Google Scholar  

Roncoroni, Miguel & Verstrepen, Kevin Joan. Belgian Beer: Tested and Tasted. (Lannoo, 2018).

Meilgaard, M. Flavor chemistry of beer: Part II: Flavor and threshold of 239 aroma volatiles. in (1975).

Bokulich, N. A. & Bamforth, C. W. The microbiology of malting and brewing. Microbiol. Mol. Biol. Rev. MMBR 77 , 157–172 (2013).

Dzialo, M. C., Park, R., Steensels, J., Lievens, B. & Verstrepen, K. J. Physiology, ecology and industrial applications of aroma formation in yeast. FEMS Microbiol. Rev. 41 , S95–S128 (2017).

Article   PubMed   PubMed Central   Google Scholar  

Datta, A. et al. Computer-aided food engineering. Nat. Food 3 , 894–904 (2022).

American Society of Brewing Chemists. Beer Methods. (American Society of Brewing Chemists, St. Paul, MN, U.S.A.).

Olaniran, A. O., Hiralal, L., Mokoena, M. P. & Pillay, B. Flavour-active volatile compounds in beer: production, regulation and control. J. Inst. Brew. 123 , 13–23 (2017).

Verstrepen, K. J. et al. Flavor-active esters: Adding fruitiness to beer. J. Biosci. Bioeng. 96 , 110–118 (2003).

Meilgaard, M. C. Flavour chemistry of beer. part I: flavour interaction between principal volatiles. Master Brew. Assoc. Am. Tech. Q 12 , 107–117 (1975).

Briggs, D. E., Boulton, C. A., Brookes, P. A. & Stevens, R. Brewing 227–254. (Woodhead Publishing). https://doi.org/10.1533/9781855739062.227 (2004).

Bossaert, S., Crauwels, S., De Rouck, G. & Lievens, B. The power of sour - A review: Old traditions, new opportunities. BrewingScience 72 , 78–88 (2019).

Google Scholar  

Verstrepen, K. J. et al. Flavor active esters: Adding fruitiness to beer. J. Biosci. Bioeng. 96 , 110–118 (2003).

Snauwaert, I. et al. Microbial diversity and metabolite composition of Belgian red-brown acidic ales. Int. J. Food Microbiol. 221 , 1–11 (2016).

Spitaels, F. et al. The microbial diversity of traditional spontaneously fermented lambic beer. PLoS ONE 9 , e95384 (2014).

Blanco, C. A., Andrés-Iglesias, C. & Montero, O. Low-alcohol Beers: Flavor Compounds, Defects, and Improvement Strategies. Crit. Rev. Food Sci. Nutr. 56 , 1379–1388 (2016).

Jackowski, M. & Trusek, A. Non-Alcohol. beer Prod. – Overv. 20 , 32–38 (2018).

Takoi, K. et al. The contribution of geraniol metabolism to the citrus flavour of beer: Synergy of geraniol and β-citronellol under coexistence with excess linalool. J. Inst. Brew. 116 , 251–260 (2010).

Kroeze, J. H. & Bartoshuk, L. M. Bitterness suppression as revealed by split-tongue taste stimulation in humans. Physiol. Behav. 35 , 779–783 (1985).

Mennella, J. A. et al. A spoonful of sugar helps the medicine go down”: Bitter masking bysucrose among children and adults. Chem. Senses 40 , 17–25 (2015).

Wietstock, P., Kunz, T., Perreira, F. & Methner, F.-J. Metal chelation behavior of hop acids in buffered model systems. BrewingScience 69 , 56–63 (2016).

Sancho, D., Blanco, C. A., Caballero, I. & Pascual, A. Free iron in pale, dark and alcohol-free commercial lager beers. J. Sci. Food Agric. 91 , 1142–1147 (2011).

Rodrigues, H. & Parr, W. V. Contribution of cross-cultural studies to understanding wine appreciation: A review. Food Res. Int. 115 , 251–258 (2019).

Korneva, E. & Blockeel, H. Towards better evaluation of multi-target regression models. in ECML PKDD 2020 Workshops (eds. Koprinska, I. et al.) 353–362 (Springer International Publishing, Cham, 2020). https://doi.org/10.1007/978-3-030-65965-3_23 .

Gastón Ares. Mathematical and Statistical Methods in Food Science and Technology. (Wiley, 2013).

Grinsztajn, L., Oyallon, E. & Varoquaux, G. Why do tree-based models still outperform deep learning on tabular data? Preprint at http://arxiv.org/abs/2207.08815 (2022).

Gries, S. T. Statistics for Linguistics with R: A Practical Introduction. in Statistics for Linguistics with R (De Gruyter Mouton, 2021). https://doi.org/10.1515/9783110718256 .

Lundberg, S. M. et al. From local explanations to global understanding with explainable AI for trees. Nat. Mach. Intell. 2 , 56–67 (2020).

Ickes, C. M. & Cadwallader, K. R. Effects of ethanol on flavor perception in alcoholic beverages. Chemosens. Percept. 10 , 119–134 (2017).

Kato, M. et al. Influence of high molecular weight polypeptides on the mouthfeel of commercial beer. J. Inst. Brew. 127 , 27–40 (2021).

Wauters, R. et al. Novel Saccharomyces cerevisiae variants slow down the accumulation of staling aldehydes and improve beer shelf-life. Food Chem. 398 , 1–11 (2023).

Li, H., Jia, S. & Zhang, W. Rapid determination of low-level sulfur compounds in beer by headspace gas chromatography with a pulsed flame photometric detector. J. Am. Soc. Brew. Chem. 66 , 188–191 (2008).

Dercksen, A., Laurens, J., Torline, P., Axcell, B. C. & Rohwer, E. Quantitative analysis of volatile sulfur compounds in beer using a membrane extraction interface. J. Am. Soc. Brew. Chem. 54 , 228–233 (1996).

Molnar, C. Interpretable Machine Learning: A Guide for Making Black-Box Models Interpretable. (2020).

Zhao, Q. & Hastie, T. Causal interpretations of black-box models. J. Bus. Econ. Stat. Publ. Am. Stat. Assoc. 39 , 272–281 (2019).

Article   MathSciNet   Google Scholar  

Hastie, T., Tibshirani, R. & Friedman, J. The Elements of Statistical Learning. (Springer, 2019).

Labrado, D. et al. Identification by NMR of key compounds present in beer distillates and residual phases after dealcoholization by vacuum distillation. J. Sci. Food Agric. 100 , 3971–3978 (2020).

Lusk, L. T., Kay, S. B., Porubcan, A. & Ryder, D. S. Key olfactory cues for beer oxidation. J. Am. Soc. Brew. Chem. 70 , 257–261 (2012).

Gonzalez Viejo, C., Torrico, D. D., Dunshea, F. R. & Fuentes, S. Development of artificial neural network models to assess beer acceptability based on sensory properties using a robotic pourer: A comparative model approach to achieve an artificial intelligence system. Beverages 5 , 33 (2019).

Gonzalez Viejo, C., Fuentes, S., Torrico, D. D., Godbole, A. & Dunshea, F. R. Chemical characterization of aromas in beer and their effect on consumers liking. Food Chem. 293 , 479–485 (2019).

Gilbert, J. L. et al. Identifying breeding priorities for blueberry flavor using biochemical, sensory, and genotype by environment analyses. PLOS ONE 10 , 1–21 (2015).

Goulet, C. et al. Role of an esterase in flavor volatile variation within the tomato clade. Proc. Natl. Acad. Sci. 109 , 19009–19014 (2012).

Article   ADS   CAS   PubMed   PubMed Central   Google Scholar  

Borisov, V. et al. Deep Neural Networks and Tabular Data: A Survey. IEEE Trans. Neural Netw. Learn. Syst. 1–21 https://doi.org/10.1109/TNNLS.2022.3229161 (2022).

Statista. Statista Consumer Market Outlook: Beer - Worldwide.

Seitz, H. K. & Stickel, F. Molecular mechanisms of alcoholmediated carcinogenesis. Nat. Rev. Cancer 7 , 599–612 (2007).

Voordeckers, K. et al. Ethanol exposure increases mutation rate through error-prone polymerases. Nat. Commun. 11 , 3664 (2020).

Goelen, T. et al. Bacterial phylogeny predicts volatile organic compound composition and olfactory response of an aphid parasitoid. Oikos 129 , 1415–1428 (2020).

Article   ADS   Google Scholar  

Reher, T. et al. Evaluation of hop (Humulus lupulus) as a repellent for the management of Drosophila suzukii. Crop Prot. 124 , 104839 (2019).

Stein, S. E. An integrated method for spectrum extraction and compound identification from gas chromatography/mass spectrometry data. J. Am. Soc. Mass Spectrom. 10 , 770–781 (1999).

American Society of Brewing Chemists. Sensory Analysis Methods. (American Society of Brewing Chemists, St. Paul, MN, U.S.A., 1992).

McAuley, J., Leskovec, J. & Jurafsky, D. Learning Attitudes and Attributes from Multi-Aspect Reviews. Preprint at https://doi.org/10.48550/arXiv.1210.3926 (2012).

Meilgaard, M. C., Carr, B. T. & Carr, B. T. Sensory Evaluation Techniques. (CRC Press, Boca Raton). https://doi.org/10.1201/b16452 (2014).

Schreurs, M. et al. Data from: Predicting and improving complex beer flavor through machine learning. Zenodo https://doi.org/10.5281/zenodo.10653704 (2024).

Download references

Acknowledgements

We thank all lab members for their discussions and thank all tasting panel members for their contributions. Special thanks go out to Dr. Karin Voordeckers for her tremendous help in proofreading and improving the manuscript. M.S. was supported by a Baillet-Latour fellowship, L.C. acknowledges financial support from KU Leuven (C16/17/006), F.A.T. was supported by a PhD fellowship from FWO (1S08821N). Research in the lab of K.J.V. is supported by KU Leuven, FWO, VIB, VLAIO and the Brewing Science Serves Health Fund. Research in the lab of T.W. is supported by FWO (G.0A51.15) and KU Leuven (C16/17/006).

Author information

These authors contributed equally: Michiel Schreurs, Supinya Piampongsant, Miguel Roncoroni.

Authors and Affiliations

VIB—KU Leuven Center for Microbiology, Gaston Geenslaan 1, B-3001, Leuven, Belgium

Michiel Schreurs, Supinya Piampongsant, Miguel Roncoroni, Lloyd Cool, Beatriz Herrera-Malaver, Florian A. Theßeling & Kevin J. Verstrepen

CMPG Laboratory of Genetics and Genomics, KU Leuven, Gaston Geenslaan 1, B-3001, Leuven, Belgium

Leuven Institute for Beer Research (LIBR), Gaston Geenslaan 1, B-3001, Leuven, Belgium

Laboratory of Socioecology and Social Evolution, KU Leuven, Naamsestraat 59, B-3000, Leuven, Belgium

Lloyd Cool, Christophe Vanderaa & Tom Wenseleers

VIB Bioinformatics Core, VIB, Rijvisschestraat 120, B-9052, Ghent, Belgium

Łukasz Kreft & Alexander Botzki

AB InBev SA/NV, Brouwerijplein 1, B-3000, Leuven, Belgium

Philippe Malcorps & Luk Daenen

You can also search for this author in PubMed   Google Scholar

Contributions

S.P., M.S. and K.J.V. conceived the experiments. S.P., M.S. and K.J.V. designed the experiments. S.P., M.S., M.R., B.H. and F.A.T. performed the experiments. S.P., M.S., L.C., C.V., L.K., A.B., P.M., L.D., T.W. and K.J.V. contributed analysis ideas. S.P., M.S., L.C., C.V., T.W. and K.J.V. analyzed the data. All authors contributed to writing the manuscript.

Corresponding author

Correspondence to Kevin J. Verstrepen .

Ethics declarations

Competing interests.

K.J.V. is affiliated with bar.on. The other authors declare no competing interests.

Peer review

Peer review information.

Nature Communications thanks Florian Bauer, Andrew John Macintosh and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary information, peer review file, description of additional supplementary files, supplementary data 1, supplementary data 2, supplementary data 3, supplementary data 4, supplementary data 5, supplementary data 6, supplementary data 7, reporting summary, source data, source data, rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Schreurs, M., Piampongsant, S., Roncoroni, M. et al. Predicting and improving complex beer flavor through machine learning. Nat Commun 15 , 2368 (2024). https://doi.org/10.1038/s41467-024-46346-0

Download citation

Received : 30 October 2023

Accepted : 21 February 2024

Published : 26 March 2024

DOI : https://doi.org/10.1038/s41467-024-46346-0

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing: Translational Research newsletter — top stories in biotechnology, drug discovery and pharma.

College & Research Libraries News  ( C&RL News ) is the official newsmagazine and publication of record of the Association of College & Research Libraries,  providing articles on the latest trends and practices affecting academic and research libraries.

C&RL News  became an online-only publication beginning with the January 2022 issue.

C&RL News  Reader Survey

Give us your feedback in the 2024  C&RL News   reader survey ! The survey asks a series of questions today to gather your thoughts on the contents and presentation of the magazine and should only take approximately 5-7 minutes to complete. Thank you for taking the time to provide your feedback and suggestions for  C&RL News , we greatly appreciate and value your input.

ALA JobLIST

Advertising Information

• Preparing great speeches: A 10-step approach (211141 views)
• The American Civil War: A collection of free online primary sources (197818 views)
• 2018 top trends in academic libraries: A review of the trends and issues affecting academic libraries in higher education (77555 views)

Scholarly Communication

Call for Paper Submissions

C&RL News Scholarly Communication Column

T he Scholarly Communication column of C&RL News is looking for papers for the June, July/August, and September 2024 issues. Edited by a subgroup of the ACRL Research and Scholarly Environment Committee, columns are not quite feature/research articles and not quite editorials, but tend to fall somewhere in between. Citations should be kept to a reasonable number. Authors can and do express their opinions in their pieces, and these opinions do not necessarily need to be shared by the column editors. The editors reserve the right to decline to publish pieces, even pieces they previously agreed to publish, for any reason. Columns will typically be in the 1,800–2,000 word range, counting citations. The absolute minimum length should be 1,500 words, and the max is 2,200 words.

Key topics covered by this column (not limited to these topics)

• Open access (OA)
• Open educational resources (OER)
• Publishing industry
• Impact of research and scholarly work (including bibliometrics and altmetrics)
• Institutional repositories
• Data repositories
• Diversity, equity, inclusion, and accessibility (DEIA) in scholarly communication
• Assessment of scholarly communication needs
• Open monographs
• Data management
• Transformative agreements
• Copyright and fair use
• Associations, groups, communities of practices related to scholarly communication

For complete details on publishing with C&RL News , visit the Instructions for Authors at https://crln.acrl.org/index.php/crlnews/about/submissions#authorGuidelines . Interested in submitting a proposal or topic for review? Please email the column editors at [email protected] .

Article Views (Last 12 Months)

Contact ACRL for article usage statistics from 2010-April 2017.

Article Views (By Year/Month)

© 2024 Association of College and Research Libraries , a division of the American Library Association

Print ISSN: 0099-0086 | Online ISSN: 2150-6698

ALA Privacy Policy

ISSN: 2150-6698

Faculty of Engineering

Quick Access

• About PolyU
• PolyU 85th Anniversary
• PolyU A to Z
• Research at PolyU
• Study at PolyU

• Academic and Teaching Staff
• Distinguished, Honorary, Emeritus, Adjunct and Visiting Staff
• Supporting Staff
• BEng(Hons) in Electrical Engineering
• BEng(Hons) in Transportation Engineering
• BEng (Hons) in Electrical Engineering with a Secondary Major in Artificial Intelligence and Data Analytics
• BEng(Hons) in Electronic Systems and Internet-of-Things
• BSc(Hons) in Artificial Intelligence and Information Engineering
• BSc(Hons) in Information Security
• BEng(Hons) in Internet-of-Things with a Secondary major in Artificial Intelligence and Data Analytics
• Engineering Doctorate in Electrical Engineering
• Master of Science in Electrical Engineering
• Master of Science in Electronic and Information Engineering
• Master of Science in Electric Vehicles
• Master of Science in Microelectronics and Quantum Systems Engineering
• Admission Requirements
• Hong Kong PhD Fellowship Scheme (HKPFS)
• Joint PhD Programmes Leading to Dual Awards
• Collaborative PhD Training Programmes
• Subject Syllabi
• Programme Documents
• Credit Transfer

Experience and Opportunities

• Engineering Entrepreneurship Club
• PolyU Capture-The-Flag Team
• Mentorship Programme
• Student Activities
• Academic Advising
• Scholarships
• Exchange Programme
• Artificial Intelligence and Signal Processing
• Power Electronics & Electric Vehicles
• Professional Services
• Research Centres and Labs
• News and Events
• Message from Head
• Departmental Academic Advisor
• Departmental Advisory Committee
• Career Opportunities

Publication of two papers on Nature Communications by Prof. Gang Li’s research team

Prof. Gang Li’s team published two papers on Nature Communications on 28 February and 2 March 2024, reporting research breakthroughs in the field of organic solar cells. In the work “Rational molecular and device design enables organic solar cells approaching 20% efficiency”, they systematically investigated the design rules of guest molecules by theoretical simulations and successfully synthesized a new molecule, o-BTP-eC9, and achieved a record organic solar cell with 19.9% (19.5 certified) efficiency. The work is in collaboration with the Chinese Academy of Science – Chongqing and the Taizhou University, Zhejiang. The work “Efficient all-small-molecule organic solar “ cells processed with non-halogen solvent ” , molecular design and control of molecular pre-aggregation has led to successful improvement of morphology in all-small-molecule organic solar cells with environmentally friendly non-halogen solvent processing. The work is in collaboration with the Huaqiao University, Xiamen and the City University of Hong Kong. Both articles are featured in Nature Communications Editors’ Highlights webpage of recent research called “Devices” Editors' Highlights | Nature Communications . The two journal papers are available at the following links. https://www.nature.com/articles/s41467-024-46022-3

https://www.nature.com/articles/s41467-024-46144-8

We use Cookies to give you a better experience on our website. By continuing to browse the site without changing your privacy settings, you are consenting to our use of Cookies. For more information, please see our Privacy Policy Statement .

Your browser is not the latest version. If you continue to browse our website, Some pages may not function properly. You are recommended to upgrade to a newer version or switch to a different browser. A list of the web browsers that we support can be found here

What are you looking for?

Popular Search

• Competition

We've detected unusual activity from your computer network

To continue, please click the box below to let us know you're not a robot.

Why did this happen?

Please make sure your browser supports JavaScript and cookies and that you are not blocking them from loading. For more information you can review our Terms of Service and Cookie Policy .

For inquiries related to this message please contact our support team and provide the reference ID below.