Cross-language Aspect Level Sentiment Analysis Based on Capsule Network
The purpose of cross-language sentiment analysis is to use source languages with abundant data resources to assist target languages with limited resources in sentiment analysis.A cross-language aspect level sentiment classification method BBCapNet based on Capsule Networks is proposed to address the issue of limited corpus for Chinese text annotation and overlapping sentiment polarity features of different aspects,which affects the accuracy of text sentiment analysis.This method uses the BERT model to learn semantic features of the source language and train word vectors as embedding layers,and then uses BiLSTM to learn contextual information,The Capsule Network is used to obtain the relationship between the local information and the overall sentiment polarity in the text,so as to extract the sentiment characteristics of different aspects.Finally,the normalized exponential function(softmax)is used for classification.By comparing with other mainstream methods,the results show that this method has a significant improvement in cross-language aspect level sentiment classification.
国家科技期刊平台
NSTL
万方数据
