Theme-Driven Keyphrase Extraction to Analyze Social Media Discourse

Social media platforms are vital resources for sharing self-reported health experiences, offering rich data on various health topics. Despite advancements in Natural Language Processing (NLP) enabling large-scale social media data analysis, a gap remains in applying keyphrase extraction to health-related content. Keyphrase extraction is used to identify salient concepts in social media discourse without being constrained by predefined entity classes. This paper introduces a theme-driven keyphrase extraction framework tailored for social media, a pioneering approach designed to capture clinically relevant keyphrases from user-generated health texts. Themes are defined as broad categories determined by the objectives of the extraction task. We formulate this novel task of theme-driven keyphrase extraction and demonstrate its potential for efficiently mining social media text for the…