Review of Public Opinion Sentiment Recognition Technology

With the rapid development of the internet and social media, public opinion sentiment recognition has become a crucial research topic. This paper first introduces the background and significance of sentiment recognition technology, followed by a comprehensive classification and analysis of single-modal and multi-modal approaches in public opinion sentiment recognition. Single-modal research primarily focuses on textual data analysis, employing techniques such as sentiment lexicons, machine learning, and deep learning for emotion classification. Multi-modal research enhances the accuracy and robustness of sentiment recognition by integrating information from diverse modalities, including text, audio, and visual data. Although existing technologies have achieved notable progress, challenges remain, such as difficulties in processing complex emotions, limited model interpretability, and insufficient generalization capabilities. Future research should prioritize multi-modal data fusion, fine-grained sentiment classification, and improvements in real-time processing and model interpretability to better support enterprise decision-making and social governance.