基于R-Boson的在线课程评论情感分析模型研究

Research on Sentiment Analysis Model for Online Course Comments Based on R-Boson

随着在线教育平台的普及,蕴含丰富情感信息的在线课程评论文本不断涌现,其对于优化在线教育平台和提升教学效果具有重要意义。故构建一种基于R-Boson情感词典的在线课程评论情感分析模型。首先,爬取B站课程评论并运用jieba等技术进行数据预处理;其次,根据评论特点建立教育领域否定词和程度副词词典;最后,使用R-Boson情感分析模型计算评论情感倾向。结果表明,与基础Boson词典相比,添加否定词和程度副词的R-Boson模型性能有所提升,其F1值从93%提升至95%,负向召回率从54%提升至79%,负向精确率从76%提升至87%;同时,模型在递增数据规模下F1值从89%逐渐提升至95%。

With the popularization of online education platforms,online course review texts containing rich emotional information continue to emerge,which is of great significance for optimizing online education platforms and improving teaching effectiveness.Therefore,a sentiment analysis model for online course comments based on R-Boson sentiment dictionary is constructed.Firstly,it crawls course comments from bilibili and uses techniques such as jieba for data preprocessing.Secondly,it establishes a dictionary of negative words and degree adverbs in the field of education based on the characteristics of comments.Finally,it uses the R-Boson sentiment analysis model to calculate the sentiment tendency of comments.The results show that compared with the basic Boson dictionary,the R-Boson model with negative words and degree adverbs improves its performance.Its F1 value increases from 93%to 95%,the negative recall rate increases from 54%to 79%,and the negative accuracy rate increases from 76%to 87%.At the same time,the F1 value of the model gradually increases from 89%to 95%in increasing data size.