Pub Date : 2022-11-03DOI: 10.1109/SMAP56125.2022.9942096
Katerina Tzafilkou, Fotini-Rafailia Panavou, A. Economides
When viewers watch food video campaigns in social media, they are experiencing various emotions. This study explores these viewers’ emotional states that can be detected through facial expressions, as well as their perceived hedonic liking of the product. The study consists of five experiments, one per product/stimulus including hedonic and utilitarian foods. Seventy-six viewers successfully participated in the tasks, and 164 valid video records were analyzed by FaceReader Online. The results indicated that FaceReader Online can capture differences in emotional responses elicited by different types of food and media posts in social media marketing campaigns. Sadness prevailed all other emotions throughout the campaigns, while arousal remained at levels of inactivity. The responses in the hedonic product campaign were significantly less negative (in terms of sadness and anger) than those in the campaigns of utilitarian products. The hedonic liking ratings indicated significant differences among campaigns of similar content and media characteristics, implying the determinant role of other factors, like individual product preferences and sensory expectations. The results contribute to understanding consumer emotions during watching food related campaigns in social media.
{"title":"Facially Expressed Emotions and Hedonic Liking on Social Media Food Marketing Campaigns:Comparing Different Types of Products and Media Posts","authors":"Katerina Tzafilkou, Fotini-Rafailia Panavou, A. Economides","doi":"10.1109/SMAP56125.2022.9942096","DOIUrl":"https://doi.org/10.1109/SMAP56125.2022.9942096","url":null,"abstract":"When viewers watch food video campaigns in social media, they are experiencing various emotions. This study explores these viewers’ emotional states that can be detected through facial expressions, as well as their perceived hedonic liking of the product. The study consists of five experiments, one per product/stimulus including hedonic and utilitarian foods. Seventy-six viewers successfully participated in the tasks, and 164 valid video records were analyzed by FaceReader Online. The results indicated that FaceReader Online can capture differences in emotional responses elicited by different types of food and media posts in social media marketing campaigns. Sadness prevailed all other emotions throughout the campaigns, while arousal remained at levels of inactivity. The responses in the hedonic product campaign were significantly less negative (in terms of sadness and anger) than those in the campaigns of utilitarian products. The hedonic liking ratings indicated significant differences among campaigns of similar content and media characteristics, implying the determinant role of other factors, like individual product preferences and sensory expectations. The results contribute to understanding consumer emotions during watching food related campaigns in social media.","PeriodicalId":432172,"journal":{"name":"2022 17th International Workshop on Semantic and Social Media Adaptation & Personalization (SMAP)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126041143","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-03DOI: 10.1109/SMAP56125.2022.9942216
Efthymia Moraitou, M. Konstantakis, Angeliki Chrysanthi, Yannis Christodoulou, G. Pavlidis, G. Caridakis
Open laboratories (OpenLabs) in Cultural Heritage (CH) institutions constitute an effective practice for providing visibility of all the processes that take place “behind the scenes”, as well as for the promotion of documentation data, which the specialists of the domain collect and produce. However, a simple “presentation” of processes, or the absence of necessary further explanation and communication with the specialists, may be problematic in terms of what visitors eventually see and understand. The exploitation of digital media and their efficient management and interlinking to meaningful data and knowledge may contribute significantly to the dissemination of publicly available information and the support of OpenLabs. Considering all the above, the CAnTi (Conservation of Ancient Tiryns) research project aims to design and implement virtual and augmented reality interactive applications that will visualize the conservation and restoration (CnR) data of the Acropolis of Ancient Tiryns. The digital content of the applications will be modeled using Semantic Web (SW) technologies, providing cultural visitors with access to insight documentation data and media produced by CnR scientists. The applications will constitute a part of the OpenLab activities that will be carried out on the archaeological site, enhancing the visitors’ experience regarding the CnR of the site’s current practices and past.
{"title":"Supporting conservation and restoration through digital media modeling and exploitation - the example of the Acropolis of Ancient Tiryns","authors":"Efthymia Moraitou, M. Konstantakis, Angeliki Chrysanthi, Yannis Christodoulou, G. Pavlidis, G. Caridakis","doi":"10.1109/SMAP56125.2022.9942216","DOIUrl":"https://doi.org/10.1109/SMAP56125.2022.9942216","url":null,"abstract":"Open laboratories (OpenLabs) in Cultural Heritage (CH) institutions constitute an effective practice for providing visibility of all the processes that take place “behind the scenes”, as well as for the promotion of documentation data, which the specialists of the domain collect and produce. However, a simple “presentation” of processes, or the absence of necessary further explanation and communication with the specialists, may be problematic in terms of what visitors eventually see and understand. The exploitation of digital media and their efficient management and interlinking to meaningful data and knowledge may contribute significantly to the dissemination of publicly available information and the support of OpenLabs. Considering all the above, the CAnTi (Conservation of Ancient Tiryns) research project aims to design and implement virtual and augmented reality interactive applications that will visualize the conservation and restoration (CnR) data of the Acropolis of Ancient Tiryns. The digital content of the applications will be modeled using Semantic Web (SW) technologies, providing cultural visitors with access to insight documentation data and media produced by CnR scientists. The applications will constitute a part of the OpenLab activities that will be carried out on the archaeological site, enhancing the visitors’ experience regarding the CnR of the site’s current practices and past.","PeriodicalId":432172,"journal":{"name":"2022 17th International Workshop on Semantic and Social Media Adaptation & Personalization (SMAP)","volume":"76 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114055384","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-03DOI: 10.1109/SMAP56125.2022.9942201
E. Stathopoulos, S. Diplaris, Anastasios I. Karageorgiadis, Alexandros Kokkalas, S. Vrochidis, Y. Kompatsiaris
The refugee crises have been considered as devastating humanitarian incidents throughout human history. They involve forced migrations due to war conflicts, diseases and so on, and are more relevant to nowadays than ever. What changed during the past decades and can be exploited towards greater good is the adoption of web and social media. In this paper, the main focus delves around smart retrieving of information from online sources, such as Twitter, YouTube and culturally-dedicated websites to provide cultural experts with relevant multimedia. The final scope is to build immersive experiences about migrant stories for local communities towards a more inclusive Europe. Moreover, semantic web technologies are deployed to homogenize multi-modal data and metadata into a unified knowledge graph including ontological structures for precise annotations. Additionally, this enables knowledge extraction and insights acquisition from implicit relationships. Finally, a system-wise benchmark for all utilities is showcased to evaluate each framework distinctly.
{"title":"Social Media and Web Sensing with Semantic Integration on the Refugee Crisis","authors":"E. Stathopoulos, S. Diplaris, Anastasios I. Karageorgiadis, Alexandros Kokkalas, S. Vrochidis, Y. Kompatsiaris","doi":"10.1109/SMAP56125.2022.9942201","DOIUrl":"https://doi.org/10.1109/SMAP56125.2022.9942201","url":null,"abstract":"The refugee crises have been considered as devastating humanitarian incidents throughout human history. They involve forced migrations due to war conflicts, diseases and so on, and are more relevant to nowadays than ever. What changed during the past decades and can be exploited towards greater good is the adoption of web and social media. In this paper, the main focus delves around smart retrieving of information from online sources, such as Twitter, YouTube and culturally-dedicated websites to provide cultural experts with relevant multimedia. The final scope is to build immersive experiences about migrant stories for local communities towards a more inclusive Europe. Moreover, semantic web technologies are deployed to homogenize multi-modal data and metadata into a unified knowledge graph including ontological structures for precise annotations. Additionally, this enables knowledge extraction and insights acquisition from implicit relationships. Finally, a system-wise benchmark for all utilities is showcased to evaluate each framework distinctly.","PeriodicalId":432172,"journal":{"name":"2022 17th International Workshop on Semantic and Social Media Adaptation & Personalization (SMAP)","volume":"156 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127857787","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-03DOI: 10.1109/SMAP56125.2022.9941877
G. Drakopoulos, Phivos Mylonas
Higher education is a major social institution with a multifaceted influence as for instance it increases the literacy and critical thinking level of the general population, it is one of the primary means of social mobility, and it provides the highly skilled personnel necessary to maintain and increase the technological momentum which is fundamental in contemporary societies. Nevertheless, the majority of the elements comprising the strategic culture of higher education have been forged with different objectives in mind. Therefore, in order for higher education institutions to remain highly relevant, a thorough review and renewal of this culture is required. This potentially radical transformation can be greatly facilitated through a set of behavioral techniques explicitly designed to encourage the shift from an outdated but familiar situation to a beneficial but unknown one. To corroborate the validity and feasibility of the proposed transition methodologies, successful applications of behavioral principles to all levels of education around the globe are provided and discussed. Moreover, evaluation metrics for assessing the results of a behavioral strategy are given.
{"title":"Facilitating Current Higher Education Trends With Behavioral Strategies","authors":"G. Drakopoulos, Phivos Mylonas","doi":"10.1109/SMAP56125.2022.9941877","DOIUrl":"https://doi.org/10.1109/SMAP56125.2022.9941877","url":null,"abstract":"Higher education is a major social institution with a multifaceted influence as for instance it increases the literacy and critical thinking level of the general population, it is one of the primary means of social mobility, and it provides the highly skilled personnel necessary to maintain and increase the technological momentum which is fundamental in contemporary societies. Nevertheless, the majority of the elements comprising the strategic culture of higher education have been forged with different objectives in mind. Therefore, in order for higher education institutions to remain highly relevant, a thorough review and renewal of this culture is required. This potentially radical transformation can be greatly facilitated through a set of behavioral techniques explicitly designed to encourage the shift from an outdated but familiar situation to a beneficial but unknown one. To corroborate the validity and feasibility of the proposed transition methodologies, successful applications of behavioral principles to all levels of education around the globe are provided and discussed. Moreover, evaluation metrics for assessing the results of a behavioral strategy are given.","PeriodicalId":432172,"journal":{"name":"2022 17th International Workshop on Semantic and Social Media Adaptation & Personalization (SMAP)","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131749423","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-03DOI: 10.1109/SMAP56125.2022.9941773
Mohammed Miaji, Yaser Miaji
Connecting with new people and expanding existing social squares is only one of the many benefits individuals may get from using social networking platforms. Social networks facilitate effective communication and cooperation, provide commercial prospects, and offer substantial social benefit. Using assumption, definition, analysis, modeling, and optimization techniques, social network issue research is productive. In this research, we categorize the known challenges of game theory applied to social networks into four categories: information dissemination, behavior analysis, community discovery, and information security. Every category may be clearly mastered in terms of knowledge application. On the basis of current research, we examine the limits of game theory and suggest future paths for social network research.
{"title":"Exploiting Game Theory Strategy and Artificial Intelligent to Analyze Social Networks: A Comprehensive Survey","authors":"Mohammed Miaji, Yaser Miaji","doi":"10.1109/SMAP56125.2022.9941773","DOIUrl":"https://doi.org/10.1109/SMAP56125.2022.9941773","url":null,"abstract":"Connecting with new people and expanding existing social squares is only one of the many benefits individuals may get from using social networking platforms. Social networks facilitate effective communication and cooperation, provide commercial prospects, and offer substantial social benefit. Using assumption, definition, analysis, modeling, and optimization techniques, social network issue research is productive. In this research, we categorize the known challenges of game theory applied to social networks into four categories: information dissemination, behavior analysis, community discovery, and information security. Every category may be clearly mastered in terms of knowledge application. On the basis of current research, we examine the limits of game theory and suggest future paths for social network research.","PeriodicalId":432172,"journal":{"name":"2022 17th International Workshop on Semantic and Social Media Adaptation & Personalization (SMAP)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124906680","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-03DOI: 10.1109/SMAP56125.2022.9941864
Theodoros Psallidas, E. Spyrou, S. Perantonis
The ever-increasing amount of user-generated audiovisual content has increased the demand for easy navigation across content collections and repositories, necessitating detailed, yet concise content representations. A typical method to this goal is to construct a visual summary, which is significantly more expressive than other alternatives, such as verbal annotations. In this paper, we describe a video summarization technique which is based on the extraction and the fusion of audio and visual data, in order to generate dynamic video summaries, i.e., video summaries that include the most essential video segments from the original video, while maintaining their original temporal sequence. Based on the extracted features, each video segment is classified as being “interesting” or “uninteresting,” and hence included or excluded from the final summary. The originality of our technique is that prior to classification, we employ a transfer learning strategy to extract deep features from pre-trained models as input to the classifiers, making them more intuitive and robust to objectiveness. We evaluate our technique on a large dataset of user-generated videos and demonstrate that the addition of deep features is able to improve classification performance, resulting in more concrete video summaries, compared to the use of only hand-crafted features.
{"title":"Summarization of User-Generated Videos Fusing Handcrafted and Deep Audiovisual Features","authors":"Theodoros Psallidas, E. Spyrou, S. Perantonis","doi":"10.1109/SMAP56125.2022.9941864","DOIUrl":"https://doi.org/10.1109/SMAP56125.2022.9941864","url":null,"abstract":"The ever-increasing amount of user-generated audiovisual content has increased the demand for easy navigation across content collections and repositories, necessitating detailed, yet concise content representations. A typical method to this goal is to construct a visual summary, which is significantly more expressive than other alternatives, such as verbal annotations. In this paper, we describe a video summarization technique which is based on the extraction and the fusion of audio and visual data, in order to generate dynamic video summaries, i.e., video summaries that include the most essential video segments from the original video, while maintaining their original temporal sequence. Based on the extracted features, each video segment is classified as being “interesting” or “uninteresting,” and hence included or excluded from the final summary. The originality of our technique is that prior to classification, we employ a transfer learning strategy to extract deep features from pre-trained models as input to the classifiers, making them more intuitive and robust to objectiveness. We evaluate our technique on a large dataset of user-generated videos and demonstrate that the addition of deep features is able to improve classification performance, resulting in more concrete video summaries, compared to the use of only hand-crafted features.","PeriodicalId":432172,"journal":{"name":"2022 17th International Workshop on Semantic and Social Media Adaptation & Personalization (SMAP)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130024304","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-03DOI: 10.1109/SMAP56125.2022.9942038
Ioannis Karamitsos, Alaa Mohasseb, Andreas Kanavos
In the modern world, social media plays a crucial role in the interchange of information and socialization with users. Twitter is a known social media platform that allows users to make relationships with others and express their opinions. The current work aims to identify the level of user engagement on Twitter with the use of graph mining. User engagement concerns the number of user connections with a tweet and can be measured using different tweet attributes including retweets, replies, etc. Specifically, this study investigates the variety of edges strength that user connections can implement in Twitter networks. Next, we employed various weights in the graph mining models to evaluate the score of each connection. These tasks were followed by statistical analysis to measure the similarity between the two user profiles as well as attributes like friendship, following and interaction in the Twitter social network. Results indicate that closely linked groups can be revealed and thus, a need for examining both group and individual behavior, will arise.
{"title":"A Graph Mining Method for Characterizing and Measuring User Engagement in Twitter","authors":"Ioannis Karamitsos, Alaa Mohasseb, Andreas Kanavos","doi":"10.1109/SMAP56125.2022.9942038","DOIUrl":"https://doi.org/10.1109/SMAP56125.2022.9942038","url":null,"abstract":"In the modern world, social media plays a crucial role in the interchange of information and socialization with users. Twitter is a known social media platform that allows users to make relationships with others and express their opinions. The current work aims to identify the level of user engagement on Twitter with the use of graph mining. User engagement concerns the number of user connections with a tweet and can be measured using different tweet attributes including retweets, replies, etc. Specifically, this study investigates the variety of edges strength that user connections can implement in Twitter networks. Next, we employed various weights in the graph mining models to evaluate the score of each connection. These tasks were followed by statistical analysis to measure the similarity between the two user profiles as well as attributes like friendship, following and interaction in the Twitter social network. Results indicate that closely linked groups can be revealed and thus, a need for examining both group and individual behavior, will arise.","PeriodicalId":432172,"journal":{"name":"2022 17th International Workshop on Semantic and Social Media Adaptation & Personalization (SMAP)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125424637","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-03DOI: 10.1109/SMAP56125.2022.9942122
Evangelia Filiopoulou, C. Bardaki, Dimitrios Boukouvalas, M. Nikolaidou, Panos E. Kourouthanassis
As consumers turn more and more to on-line shopping, requirements such as on time delivery and delivery cost saving are of major importance. Retail and logistics companies struggle to find strategies that offer a successful and fast last-mile delivery service that satisfies consumers’ preferences and expectations. Last-mile delivery is an opportunity, as well as a challenge for e-commerce retailers and logistics companies because it needs to satisfy customers’ preferences, offering the best customer experience. This paper explores Greek consumers’ preferences of last-mile delivery alternatives. We investigate the potential of using drone delivery and the challenges the consumers face during tracking their order delivery. We conducted a survey with 174 participants exploring their online shopping behavior, which delivery options they prefer, what delivery challenges they face (e.g. home delivery without prior notice) and which factors influence their delivery decisions (e.g. delivery time-flexibility and time saving). Keywords: Last-mile delivery, Drones, pick-up point
{"title":"Last-Mile Delivery Options: Exploring Customer Preferences and Challenges","authors":"Evangelia Filiopoulou, C. Bardaki, Dimitrios Boukouvalas, M. Nikolaidou, Panos E. Kourouthanassis","doi":"10.1109/SMAP56125.2022.9942122","DOIUrl":"https://doi.org/10.1109/SMAP56125.2022.9942122","url":null,"abstract":"As consumers turn more and more to on-line shopping, requirements such as on time delivery and delivery cost saving are of major importance. Retail and logistics companies struggle to find strategies that offer a successful and fast last-mile delivery service that satisfies consumers’ preferences and expectations. Last-mile delivery is an opportunity, as well as a challenge for e-commerce retailers and logistics companies because it needs to satisfy customers’ preferences, offering the best customer experience. This paper explores Greek consumers’ preferences of last-mile delivery alternatives. We investigate the potential of using drone delivery and the challenges the consumers face during tracking their order delivery. We conducted a survey with 174 participants exploring their online shopping behavior, which delivery options they prefer, what delivery challenges they face (e.g. home delivery without prior notice) and which factors influence their delivery decisions (e.g. delivery time-flexibility and time saving). Keywords: Last-mile delivery, Drones, pick-up point","PeriodicalId":432172,"journal":{"name":"2022 17th International Workshop on Semantic and Social Media Adaptation & Personalization (SMAP)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132333507","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-03DOI: 10.1109/SMAP56125.2022.9941983
Elias Dritsas, M. Trigka, Gerasimos Vonitsanos, Andreas Kanavos, Phivos Mylonas
As the volume of data generated and stored on a daily basis is constantly increasing, the need for finding techniques in terms of the automated discovery of information from them has arisen. This purpose can be effectively solved with the use of text mining, which uses methods derived from data mining, information retrieval, machine learning, as well as natural language processing. This paper addresses the problem of extracting textual information from large collections of documents by efficiently exploiting clustering techniques in a cloud computing infrastructure. The clustering was performed using three different algorithms, namely k-Means, Bisecting k-Means, and Gaussian Mixture Model (GMM). To evaluate the quality of these methods, we experimented in the Apache Spark distributed environment, on several well-known datasets, the documents of which have been manually clustered.
{"title":"An Apache Spark Implementation for Text Document Clustering","authors":"Elias Dritsas, M. Trigka, Gerasimos Vonitsanos, Andreas Kanavos, Phivos Mylonas","doi":"10.1109/SMAP56125.2022.9941983","DOIUrl":"https://doi.org/10.1109/SMAP56125.2022.9941983","url":null,"abstract":"As the volume of data generated and stored on a daily basis is constantly increasing, the need for finding techniques in terms of the automated discovery of information from them has arisen. This purpose can be effectively solved with the use of text mining, which uses methods derived from data mining, information retrieval, machine learning, as well as natural language processing. This paper addresses the problem of extracting textual information from large collections of documents by efficiently exploiting clustering techniques in a cloud computing infrastructure. The clustering was performed using three different algorithms, namely k-Means, Bisecting k-Means, and Gaussian Mixture Model (GMM). To evaluate the quality of these methods, we experimented in the Apache Spark distributed environment, on several well-known datasets, the documents of which have been manually clustered.","PeriodicalId":432172,"journal":{"name":"2022 17th International Workshop on Semantic and Social Media Adaptation & Personalization (SMAP)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132430565","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The ongoing pandemic moved many classes online, and disrupted the classroom teaching experience and the feedback loop between teachers and students. One key challenge is to detect the engagement and other affective states exhibited by students during online learning. This paper investigates the capabilities and limitations of neural networks to distinguish between different affective states (i.e., boredom, engagement, confusion, and frustration), and their intensity level (i.e., very low, low, high, and very high). Several models are built using a hybrid ResNet+TCN neural network architecture. The models are trained using a large dataset, DAiSEE, that contains short 10 second video recordings of students as they watch educational content ‘in the wild’. A second dataset consisting of longer videos, EmotiW2020, is used to cross validate the engagement level classification model. The affective state classification model outperforms prior models. Boredom, confusion and frustration level classification models outperform or are on par with prior models. The engagement level classification model achieved similar performance with other baseline models and was outperformed by some SOTA models, but those models used 5 times more frames and 5 to 10 times more training epochs. The engagement level classification model was validated and achieved similar performance on both the DAiSEE and EmotiW2020 datasets.
{"title":"Classification of Student Affective States in Online Learning using Neural Networks","authors":"Kishan Kumar Bajaj, Ioana Ghergulescu, Arghir-Nicolae Moldovan","doi":"10.1109/SMAP56125.2022.9942163","DOIUrl":"https://doi.org/10.1109/SMAP56125.2022.9942163","url":null,"abstract":"The ongoing pandemic moved many classes online, and disrupted the classroom teaching experience and the feedback loop between teachers and students. One key challenge is to detect the engagement and other affective states exhibited by students during online learning. This paper investigates the capabilities and limitations of neural networks to distinguish between different affective states (i.e., boredom, engagement, confusion, and frustration), and their intensity level (i.e., very low, low, high, and very high). Several models are built using a hybrid ResNet+TCN neural network architecture. The models are trained using a large dataset, DAiSEE, that contains short 10 second video recordings of students as they watch educational content ‘in the wild’. A second dataset consisting of longer videos, EmotiW2020, is used to cross validate the engagement level classification model. The affective state classification model outperforms prior models. Boredom, confusion and frustration level classification models outperform or are on par with prior models. The engagement level classification model achieved similar performance with other baseline models and was outperformed by some SOTA models, but those models used 5 times more frames and 5 to 10 times more training epochs. The engagement level classification model was validated and achieved similar performance on both the DAiSEE and EmotiW2020 datasets.","PeriodicalId":432172,"journal":{"name":"2022 17th International Workshop on Semantic and Social Media Adaptation & Personalization (SMAP)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117326539","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：