Graph Learning Based Sentiment Analysis System for Chinese Course Evaluation

Abstract

Natural language processing (NLP) is an important research direction of artificial intelligence. Text sentiment classification in NLP is a compromising method to exploit the constructive feedback to improve teaching quality. This paper captures the course reviews from online learning platform China University MOOC as the dataset, and uses an aspect-level sentiment classification method to analyze the course evaluation, via a graph convolution network (GCN) to characterize the syntactic dependency between context words and various aspects of sentences, and decide the emotions described by multiple non-adjacent Chinese words. As for the 1837 comments of online courses, there is obvious aggregation in the aspect of extraction. Most of the comments mainly focus on the two aspects of course and teacher, and a few comments describe other aspects related to the course. The results demonstrate that the accuracy of the model is more than 80%. Additionally, a visual interface is designed to provide the sentiment analysis results no matter what data set of course reviews is given, and make the graph learning based sentiment analysis tool user-friendly.