Research On Sentiment Analysis Of Chinese MOOC Online Comments

With the rapid development of "Internet+education",the Massive Open Online Course(MOOC)platform has also received unprecedented attention,more and more users are learning courses through the MOOC platform and leaving lots of subjective and emotional comments that contain learners' attitudes and evaluations for the courses and MOOC platform in the course comment area.It is very important for learners and administrators of MOOC platform to analyze and process these course comments efficiently,and obtain some valuable information from it.This paper takes the course comments of MOOC platform as the starting point and conducts sentiment polarity classification and potential topic mining research,the results of the classification and topic mining not only help learners to judge and analyze the quality of the courses to be selected,but also contribute to construction and development of the MOOC platform.This paper takes the Chinese University MOOC platform as an example.It uses Python web crawler technology to crawl the comments of online course learners in the comment area,and then uses deep learning and LDA-based topic model methods to conduct sentiment analysis research on learner's comments.Specific research work includes the following two parts:First,a multi-channel convolutional neural network and bidirectional gated recurrent network based on attention mechanism(MC-AttCNN-AttBiGRU)model is proposed to classify sentiment polarity of MOOC comments.Because the traditional CNN model is excellent in the local feature extraction of text,but it cannot effectively extract the contextual semantic features of the text,that is,it ignores the contextual semantic information between words,so this paper combines the bidirectional GRU and CNN model to extract the contextual semantic features of the text,which makes up for the defect of CNN,it not only can extract the local features of text comments,but also can extract the contextual semantic features of text.The introduction of attention mechanism on CNN and bidirectional GRU which allows the model automatically select the important words for text classification,ignoring the unimportant "noise" words,which is to further improve the accuracy of the sentiment classification in MOOC comments.The experimental results show that the method proposed in this paper is superior to several other machine learning methods in the sentiment polarity classification of the MOOC comments.Second,the CBOW model and the LDA topic model(CBOW-LDA)are used to mine and analyze sentiment topics in course comments of the MOOC platform,this method is a more fine-grained sentiment analysis method compared to sentiment polarity classification,it can mine the learner's attention from the perspective of potential topics in the comments and extracts the valuable information from comments for administrators of the MOOC platform.The LDA topic model has problems of inaccurate topic extraction and low processing efficiency when processing largecapacity data.In view of the above problems,this paper first uses CBOW to reduce the dimensionality of the course comments corpus,and then uses the dimensionalityreduced corpus as the input of the LDA topic model,and then uses the Gibbs sampling method to obtain the document-topic distribution and topic-word distribution to mine the potential topics and topic words in the course comments,and finally sentiment analysis experiment is carried out.Experimental results show that the method proposed in this paper is better than the traditional LDA model,Skip-gram-LDA and TF-IDFLDA model in the effect of topic extraction for the MOOC comments.