Hierarchical Semantic Sentiment‐Aware Linguistic Network for Depression Detection in Social Media

ABSTRACT

Depression is a global mental health problem, and early identification from social media text is crucial for timely intervention. However, the implicit and context‐dependent nature of depression cues in text renders their detection particularly difficult. To address these issues, this paper proposes the Hierarchical Semantic Sentiment‐aware Linguistic Network (HSSLN) for text‐based depression detection. The Hierarchical Semantic Module (HSM) applies multi‐granularity interaction to capture hierarchical semantic dependencies. The Sentiment‐aware Module (SAM) constructs a sentiment‐related matrix to model subtle sentiment cues under weak supervision. The Linguistic‐enhanced Fusion Module (LEFM) employs hierarchical modality integration to combine linguistic patterns with semantic‐sentiment representations. Experiments on the Twitter and Reddit datasets demonstrate that HSSLN effectively identifies subtle depression‐related cues by jointly leveraging semantic, sentiment and linguistic information.