Text Sentiment Analysis Based on BERT-TextCNN-BILSTM

Abstract

AbstractAs a critical component of natural language processing, text sentiment analysis has taken on an increasingly important role in areas such as recommendation systems, customer evaluation analysis and opinion referencing. In order to more accurately extract the sentiment embedded in text from text information, this paper proposes a BTB model, which is a text sentiment classification model incorporating BERT, TextCNN, and BILSTM. In the BTB model, the generated word vector has more a priori information, and the important features of the text and contextual information are efficiently used in the sentiment analysis. The results of experiments show that the BTB model achieves high precision, recall and F1 value, and has good generalisation ability.KeywordsNatural language processingText sentiment classificationBERTTextCNNBILSTM