Document-Level Sentiment Analysis of Course Review Based on BG-Caps  

作　　者：Jing Wu Tianyi Liu Wei Hu 

机构地区：[1]College of Computer Science and Technology,Wuhan University of Science and Technology,Wuhan,China

出　　处：《国际计算机前沿大会会议论文集》2022年第2期394-405,共12页International Conference of Pioneering Computer Scientists, Engineers and Educators（ICPCSEE）

摘　　要：With the development of the Internet in various fields,the combination of education and the Internet is close.Many users choose courses they are interested in to study on the MOOC platform and leave text reviews with emotional colors.However,the traditional word vector representation method extracts text information in a static way,which ignores text location information.The convolutional neural network cannot fully utilize the semantic features and correlation information,so the results of text sentiment analysis are inaccurate.To solve the above problems,this paper proposes a sentiment analysis method based on BGCaps MOOC text review.The ALBERT pretraining model was used to obtain the dynamic feature of the text.Combined with the BiGRU and capsule network model,the features were trained to obtain deep semantic features.We evaluated our mode on theMOOCreviewdataset.The results showthat the proposed method achieved effective improvement in accuracy.

关 键 词：Emotional analysis Capsule network MOOC ALBERT 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]