Paper Type

Short

Paper Number

PACIS2026-1800

Description

Traditional Chinese Medicine (TCM) social media discourse represents a high-value yet demanding corpus for digital health informatics, capturing public health discourse spanning symptom-level experiences, institutional legitimacy debates, and broader cultural interpretive frameworks. Existing topic modeling approaches often fail to recover layered semantic structures due to fine-grained topic cluster fragmentation, loss of low-frequency meaningful clusters, and lexical collapse during topic merging. This study proposes SC-BERTopic, a Semantic-Centroid Hierarchical BERTopic framework that introduces semantic centroids in embedding space to preserve semantic-space consistency during hierarchical topic consolidation. Applied to 666,230 Douyin comments, SC-BERTopic outperforms standard BERTopic across coherence, structural balance, and human evaluation. The identified three-tier thematic hierarchy reveals how TCM-related public discourse becomes hierarchically organized from symptom concerns to institutional legitimacy debates and value-oriented discourse framing. This study contributes to IS research on digital health sensemaking while providing a transferable framework and open benchmark corpus for computational health communication research.

Comments

14-Healthcare

Download

Share

COinS
 
Jul 5th, 12:00 AM

SC-BERTopic: Semantic-Centroid Hierarchical Modeling of Traditional Chinese Medicine Social Media Discourse

Traditional Chinese Medicine (TCM) social media discourse represents a high-value yet demanding corpus for digital health informatics, capturing public health discourse spanning symptom-level experiences, institutional legitimacy debates, and broader cultural interpretive frameworks. Existing topic modeling approaches often fail to recover layered semantic structures due to fine-grained topic cluster fragmentation, loss of low-frequency meaningful clusters, and lexical collapse during topic merging. This study proposes SC-BERTopic, a Semantic-Centroid Hierarchical BERTopic framework that introduces semantic centroids in embedding space to preserve semantic-space consistency during hierarchical topic consolidation. Applied to 666,230 Douyin comments, SC-BERTopic outperforms standard BERTopic across coherence, structural balance, and human evaluation. The identified three-tier thematic hierarchy reveals how TCM-related public discourse becomes hierarchically organized from symptom concerns to institutional legitimacy debates and value-oriented discourse framing. This study contributes to IS research on digital health sensemaking while providing a transferable framework and open benchmark corpus for computational health communication research.