Sentiment Analysis of Douban Book Reviews Based on BiLSTM-Attention-CNN Model

In this paper, we present a hybrid model that combines bi-directional long short-term memory networks (BiLSTM), an attention mechanism, and convolutional neural networks (CNN) for the purpose of sentiment analysis of Douban Book reviews. The proposed model effectively captures contextual information through the utilization of BiLSTM, highlights significant sentiment words via the attention mechanism, and extracts local features in conjunction with the CNN, thereby facilitating accurate classification of review sentiment. Experimental results indicate that our model achieves an accuracy of 92.3% on the test set, accompanied by an F1 score of 92.1%, surpassing the performance of both individual models and other hybrid models. The model's efficacy is further corroborated through comparisons with traditional models, such as Support Vector Machines (SVM) and BERT. This research contributes valuable insights to the field of sentiment analysis and offers practical applications for book recommendations and the analysis of reader feedback.