Chinese Sentiment Analysis of MOOC Reviews Based on Word Vectors

Abstract

Sentiment analysis of MOOC reviews can reflect learners’ learning feelings, topics of interest and satisfaction with courses to platform managers and teachers, so as to help them make improvements and improve the quality of teaching. Potential learners can also find out the emotional information carried by the attributes of the courses they care about according to the reviews, which can be used as the basis for choosing courses or not. From the perspective of word vectors, this paper uses two word vector tools, Word2vec and Glove, to train MOOC review word vector data respectively, compares the impact of the two word vector tools on sentiment analysis, and visualizes the weight of attention. The experiment found that Word2vec was better than Glove on the whole, and the CBOW model under Word2vec had the best performance in binary classification, while the skip-gram model had the best performance in three classification. In addition, the effects of the two tools in finding similar words in new words was discussed. The experiment found that the Word2vec was better than the Glove, and the skip-gram model was more accurate than the CBOW model.