Aνάλυση συναισθήματος κατά τη διάρκεια εξ αποστάσεως εκπαίδευσης σε εφαρμογή Android και αποστολή σε RESTful Web Service
Sentiment analysis during online learning on Android application and sending to a RESTful Web Service
View/
Keywords
Android ; Tensorflow Lite ; Spring Boot ; Mηχανική μάθηση ; Συνελικτικό νευρωνικό δίκτυο ; Ανάλυση συναισθήματος προσώπου ; RESTful web service ; Java ; Python ; PostgreSQLAbstract
This thesis is about an Android application that predicts facial emotions of university students during distance learning. Specifically, the application simulates an online lecture, during which it periodically performs facial emotion recognition on snapshots of students who gave the required permission to the application and sends the algorithm’s prediction from their online devices, who act as clients, to a database through a web service. From the contents of the database, quantitative data are created, which can be accessed by the teaching professor, enabling him to draw conclusions about the impact his style of teaching made either at a specific lecture or during the course of a semester. An implementation like this, after getting expanded, could be used either independently or as an add-on to widely used platforms, to help insure the quality of online learning by providing the professor with almost immediate feedback on the students’ response to the content and the method of his teaching.
In the following pages an overview is being provided of neural networks, and specifically a subcategory of theirs, the convolutional neural network, a model of which was used for predicting the emotion of the students using the application, of the Tensorflow Lite library, which allows the deployment of machine learning models on an Android device and of the Spring Boot Framework, with which the RESTful web service for the application was created. Subsequently, a presentation is being given on the process of training the neural network, designing the database and developing the application. Lastly, a mention is being made of conclusions which came up after completing the implementation and of probable future expansions.