| With the rapid development of mobile learning technology,MOOC has become a new platform for online learning.The comments section of a MOOC contains students’ opinions and suggestions about the course,and the analysis of comment data is of great value in improving the quality of teaching.Currently,research on comment data analysis can provide some information about the quality of online courses or students’ learning behavior to a certain extent,but cannot obtain main opinions of students about the course(i.e.,hot opinions),nor can it improve classroom teaching effectiveness in a targeted manner.To address the shortcomings of existing research,this thesis analyzes and studies the sentiment in MOOC comments and proposes a sentiment analysis method to address the two aforementioned issues.The main research contents and innovations of this thesis are as follows:(1)This thesis proposes an Extraction Method for Hot Opinions based on MOOC comments(EMHO),which solves the problem of existing research methods not being able to summarize the specific opinions that students hold about the course based on comments.EMHO divides the process of extracting hot opinions into three stages:entity target extraction,opinion extraction,and hot opinion acquisition.In the entity target extraction stage,EMHO uses a target extraction module composed of BERT and target word filtering algorithms to obtain the entity target of course comments.In the opinion extraction stage,EMHO inputs the entity target into an opinion extraction module to obtain opinions about the entity target.In the hot opinion acquisition stage,the semantic clustering of target-opinion pairs obtained in the previous stage is performed using the K-means algorithm to obtain hot opinions.Experimental results show that EMHO can accurately extract hot opinions from MOOC comments.(2)This thesis proposes a MOOC Comments-based Constructive suggestion Extraction(MCE)method,which matches the suggestions in the comments with the entity targets from a semantic relevance perspective,solving the problem of existing research methods not being able to quickly extract suggestions from course comments and improve course quality in a targeted manner.MCE includes two steps: target and suggestion extraction,and relevance matching for target words and suggestions(RMTS)algorithm.In the target and suggestion acquisition step,the target extraction method used in EMHO is employed to extract targets and suggestions from the comments.In the target and suggestion relevance matching step,the Relevance Matching for Target words and Suggestions(RMTS)algorithm is proposed.The RMTS algorithm divides the matching process into two steps: semantic understanding and relevance decision-making.In the semantic understanding step,the input targets and suggestions are first subjected to low level semantic understanding using an improved Variational Autoencoder(VAE)model to obtain their low level semantic representations.Then,the semantic abstraction module is utilized to derive the similarity between targets and suggestions based on their low level semantic representations,as well as the higher-level semantics to which they belong.In the relevance decision-making step,based on the results obtained from the semantic abstraction module and the higher-level semantic tree,semantic relevance matching decisions are made for the targets and suggestions.Experimental evaluation of the RMTS algorithm is conducted on a MOOC dataset.The results demonstrate that the algorithm achieves significantly higher matching precision for targets and suggestions compared to other baseline models. |
