| Today,with the rise of the "Internet plus education" model,the online course platform has more and more users,what follows is the generation of a large amount of course review data.Through analysis and research of online course reviews,can understand the user’s most real feelings and thoughts.Online course comment analysis can play an important role in the rapid and accurate course promotion of platform users,feedback on the teaching quality of the instructor’s course,and the sustainable development of the platform,therefore,the research and discussion on the analysis of online course reviews becomes very meaningful.Although text sentiment analysis in various fields has developed rapidly,however,it is not very suitable for sentiment analysis of online course reviews.There are mainly the following problems: 1.Online course reviews have the characteristics of large amount of data,no fixed format for reviews,fast updates,polysemy and so on,the current sentiment analysis model is not well suited for online course reviews;2.The traditional model has simple structure,when training the word vector of comment information,it is easy to lose the original semantics of the comment information,Cause the final online course review sentiment analysis results to be inaccurate;3.Existing topic sentiment analysis models,the particularity of online course review texts is not well considered,and the topics uncovered are not covered.Therefore,in order to solve these three problems,this paper has done the following work:1.For the problem of semantic loss of comment information,this paper designs an online course review sentiment analysis model based on ALBERT-Bi-LSTM.First,use ALBERT to dynamically generate word vectors while incorporating local semantic information;Then,the global semantic information vector of the pre-order and post-order is captured by Bi-LSTM,and the attention mechanism is integrated to judge the different weights of sentences where different words are located;Finally,input softmax for sentiment classification result output.Finally,this paper conducts experiments on real online course reviews,the experimental results show that the designed model has great value in sentiment analysis of online course reviews.2.The subject emotion for extracting online course reviews needs to have certain coverage and representation.Therefore,this paper chooses to adopt the LDA topic model,It can associate words that do not seem to have any relationship through the hidden topic relationship,in this way,the potential theme sentiment is analyzed,and the excavated theme has more coverage,since it uses a bag-of-words model,the process cannot consider the order between words,in order to solve the above problems,this paper integrates the TextRank algorithm based on the LDA topic model.The TextRank algorithm constructs a course comment network node graph,extracts text order information and semantic information,therefore,the topic model is combined with the TextRank algorithm for topic sentiment analysis.Finally,this paper conducts experiments on a real online course review dataset,and the results show that the method is feasible and superior to traditional machine learning and some deep learning methods. |
PDF Full Text Request