Uma abordagem computacional para detectar emoções de alunos em cursos online

Abstract

Each year the number of educational institutions that use Virtual Learning Environments in distance learning increases. These Virtual Learning Environments generate a large volume of information and are of utmost importance for the institution, as well as for teachers, and with this information, we can identify various students’ emotions, such as frustration, isolation, discouragement, and demotivation, and through techniques of Artificial Intelligence it is possible to stimulate and motivate students. Affection is a great "ally" to promote learning and affection in virtual learning environments, the mediator's interactive actions must be more constant, being present and attentive to the students' "movement", seeking to help them individually and consciously, giving them feedback, making them feel safe and helping to cheer them up. The interaction between people in face-to-face courses obtains excellent results, as the senses help students in the execution of activities, however, in Virtual Learning Environments this perception is different and needs to be stimulated. Thus, this Thesis describes a model for automatic identification of emotions in texts produced by students of Distance Education in Virtual Learning Environments - VLEs. The approach proposed in this work allows to identify the sentiment contained in text messages of students of distance courses using automatic classification algorithm using sentiment analysis based on lexical approach. The main results obtained were that the proposed approach (lexical, based on Sentiment Analysis) can contribute positively to the teaching-learning process, considering that knowing the emotional state of the students can improve the monitoring of the class, because, with this additional information, the tutor has one more resource for decision makings, such as recognizing students with dissatisfaction and lack of interest in the course. Finally, as a limitation of the research, we have: subjectivity in the texts, production of large texts, sample size, classification time.