Research on Public Opinion Hotspot Theme Identification and Guidance Strategies Based on Sentiment Analysis and Semantic Mining

Abstract

Social media environments generate large-scale, rapidly evolving public opinion data, making accurate identification of hotspot themes and effective guidance strategies increasingly important. This study proposes a comprehensive analytical framework integrating fine-grained sentiment analysis and semantic mining to reveal the underlying thematic structures of public opinion evolution. A multi-source data acquisition and preprocessing module is first established to collect heterogeneous social media content. A hybrid sentiment analysis model combining lexicon-based methods and machine learning is then employed to quantify topic-level emotional intensity and sentiment dynamics. Furthermore, semantic network construction and community detection algorithms are utilized to identify latent hotspot clusters and characterize their evolutionary patterns. Based on the multidimensional relationships among topics, sentiment trends, and user roles, a hierarchical guidance strategy generation mechanism is developed. Experimental results demonstrate the effectiveness of the proposed framework in hotspot identification and dynamic trend analysis. The study provides methodological support for large-scale information mining and offers references for information propagation analysis, social sensing systems, and intelligent communication networks.