Sentiment Analysis of Chinese Comments using CNN Combined with BiLSTM Model based on Attention Mechanism

Abstract

This article proposes a Chinese comment sentiment analysis model based on attention mechanism, which combines CNN with BiLSTM, to address the issue of insufficient utilization of contextual information in text sentiment analysis. This model captures emotional words and contextual information in comment texts by combining CNN with BiLSTM structure, and then uses attention mechanism to weight important information in comments. And comparative experiments were conducted on two Chinese comment datasets (one takeout comment dataset and one product comment dataset) to verify the advantages of this model compared to several commonly used deep models. At the same time, ablation experiments were conducted to verify the impact of different modules of the model on this model and the effectiveness of the proposed combination mode.