Abstract:In response to the shortcomings of traditional single-machine learning-based sentiment analysis methods in feature extraction and semantic understanding, a novel e-commerce sentiment analysis model was developed. This model integrated a combination of CNN-BiLSTM and multi-head self-attention mechanisms, aiming to better address long-distance dependencies in the text and captured the semantic relationships of emotional information. This enhanced the model's robustness and generalization capabilities, consequently improving merchants' understanding of consumer sentiments and the accuracy of evaluations.Experiments were conducted on a publicly available Chinese e-commerce dataset, and the model was compared with other existing models. The experimental results indicate that this model outperforms others in terms of precision, accuracy, recall, and F1 score, among other metrics.