Research On Emotional Tendency Of Online Course Review Text For MOOC

With the development of computer and Internet technology,courses have been greatly improved under the promotion of the "Internet + education" model.High-quality educational resources have been shared,which has improved students’ learning of courses in different fields and enriched the methods of teaching development.However,many courses are also mixed with courses of low quality and low efficiency.The evaluation of the quality of MOOC courses cannot simply apply traditional course quality evaluation indicators and methods.Experts suggest that a large amount of teaching data should be used for data analysis.Exploring the factors affecting the quality of the course Based on this,this article starts from the diversified theory of course quality evaluation,and conducts research on the analysis and extraction of MOOC course comment indicators.This article attempts to comprehensively apply a variety of sentiment analysis and other artificial intelligence technologies to the research of online course quality evaluation,analyze the sentiment tendency of the student review texts of China’s MOOC,and extract online course quality evaluation indicators.The research content mainly includes three parts:First,in response to the current open source Chinese online course review sentiment analysis data set is scarce,the amount of data,and the quality of the data are poor,a 10,000-level sentiment analysis data set suitable for education and teaching research has been established through a semi-machine and semi-manual method.Second,for the purpose of extracting highly concentrated online course quality evaluation indicators with obvious keyword features,the BERT model is applied to sentiment analysis,and a topic extraction method based on BERT and LDA-Text Rank is proposed.Use the BERT model to train to obtain the word vectors of the online reviews of Mukenet,obtain the topic distribution of candidate keywords through the LDA topic model,and combine the Text Rank algorithm to calculate the relational words of each topic to generate the corresponding course quality evaluation indicators.Then combine the extracted results with the online course quality evaluation indicators issued by the Ministry of Education to expand the existing online evaluation indicators.The results show that the model scale and the amount of parameters are greatly reduced.Third,the two-way GRU online course evaluation model using BERT fusion attention mechanism.Input the acquired word features into the Bi-GRU network to realize the emotional feature extraction of MOOC comments,and combine the attention mechanism to enable the model to select the parameter weights that are more important to the text classification task and input them into the Softmax logistic regression algorithm to realize the curriculum Quality Evaluation.Through the analysis of new multi-dimensional indicators,further research has been done on student emotions.And the corresponding experiment was designed to verify the effectiveness of this indicator.The F1 value of the model in this paper reaches 92.45%,which is better than other mainstream emotional orientation analysis models,which proves the effectiveness of the method.