Research on sentiment classification for netizens based on the BERT-BiLSTM-TextCNN model

• DOI: 10.7717/peerj-cs.1005
• Publication Date: 2022-06-08T08:35:07Z
• Authors: Xuchu Jiang; Chao Song; Yucheng Xu; Ying Li; Yili Peng
• Keywords: Sentiment Analysis

Sentiment analysis of netizens’ comments can accurately grasp the psychology netizens and reduce risks brought by online public opinion. However, there is currently no effective method to solve problems short text, open word range, sometimes reversed order in comments. To better above problems, this article proposes a hybrid model sentiment classification, which based on bidirectional encoder representations from transformers (BERT), long short-term memory (BiLSTM) text convolution neural network (TextCNN) (BERT-BiLSTM-TextCNN). The experimental results show that (1) proposed combine advantages BiLSTM TextCNN; it not only captures local correlation while retaining context information but also has high accuracy stability. (2) BERT-BiLSTM-TextCNN extract important emotional more flexibly achieve multiclass classification tasks emotions accurately. innovations study are as follows: use BERT generate vectors prior full combination contextual semantics; model, mechanism obtain well; (3) TextCNN features well problem combined effect three modules significantly improve multilabel classification.