Go to ACL ARR 2025 February homepageDRMD: Explainable Depression Detection Based on Metaphorical Conceptual Mapping
Abstract: Depression is a pervasive mental health issue affecting millions globally, and social media has become a key platform for individuals to express their emotional struggles. The textual content shared by individuals with depression contains valuable insights into their mental states, yet analyzing such data presents challenges due to the complexity of indirect expressions, including metaphors. These metaphorical expressions can provide crucial insights into the psychological states of individuals with depression and play an important role in therapeutic contexts. This paper addresses the challenge of detecting depression by leveraging metaphorical information. We introduce a novel, publicly available Depression-related metaphor dataset (DRMD), which contains social media posts from individuals with depression, along with metaphorical labels and conceptual source domain mappings. This dataset is used to fine-tune large language models (LLMs), integrating metaphorical features to enhance the models’ depression detection performance. Our results demonstrate that the fine-tuned models with metaphorical information not only improve detection accuracy but also generate high-quality explanations for detection outcomes, utilizing metaphorical expressions to offer deeper insights into the mental states of individuals. This work highlights the potential of metaphorical analysis in mental health diagnostics and provides a foundation for future research in automated depression detection and explanation generation. The dataset is publicly available.
Paper Type: Long
Research Area: Linguistic theories, Cognitive Modeling and Psycholinguistics
Research Area Keywords: Depression detection,Metaphorical conceptual mapping,Mental health diagnostics
Languages Studied: English
Submission Number: 4711
Loading