Sentiment Analysis Based on Multiscale Convolutional Neural Network and Bidirectional Long Short-Term Memory

Abstract

With the prevalence of social media, Weibo has emerged as a crucial platform for individuals to express emotions and opinions, making accurate analysis of its sentiment content significantly meaningful. In order to more effectively capture the semantic and emotional information within Weibo text, a deep learning model that integrates multiscale CNN and BiLSTM is proposed. Firstly, this research utilizes a CNN model with multiscale convolutional kernels to perform feature extraction on Weibo text using various kernel sizes, thereby capturing both local and global information within the text. Subsequently, the incorporation of BiLSTM networks as a component of sequence modeling effectively captures long-distance dependencies within Weibo text. Through the combination of CNN and BiLSTM, the model gains a better understanding of the semantic structure and emotional expressions within Weibo text. Comparative experiments demonstrate that the proposed approach outperforms other sentiment analysis methods, showcasing higher accuracy and better overall effectiveness. This study proposes an innovative model for Chinese microblog sentiment analysis by integrating multi-scale CNN and BiLSTM technologies. It not only enhances the accuracy and efficiency of sentiment recognition but also deepens the theoretical foundation of online sentiment research, exerting significant influence and potential application value in fields such as social media sentiment monitoring and public opinion mining.