Modeling the Interplay of Time, Space, Users, Topics, and Sentiment: A Novel Analytical Framework for Social Media Data Analysis

This study addresses the gap in existing social media discourse research, which often treats time, space, users, topics, and sentiment in isolation. It proposes an integrated analytical framework to capture the interplay among these five dimensions, using an online controversy involving Nongfu Spring as an illustrative case to demonstrate the framework’s applicability. A multi-method approach combines temporal, spatial, and user classification analysis; deep learning-based sentiment analysis via a fine-tuned Chinese-BERT-wwm-ext model; and influential keyword identification using Graph Attention Network and improved TextRank algorithms. The framework is applied to 12,674 Weibo posts related to this controversy. Posts clustered temporally around key discussion periods, spatially concentrated in economically developed regions, and, while ordinary users produced the majority of posts, verified institutional and high-influence individual accounts became increasingly prominent in the later phases of the discussion. The integrated analysis reveals joint patterns across dimensions, such as the marked shift in user composition over the course of the crisis, that single-dimension analyses alone would not reveal. The study contributes a reusable framework for the multidimensional analysis of online discourse and illustrates its applicability through a single case; its generalizability across platforms and event types remains to be tested in future research.