This research paper discusses the use of natural language processing (NLP) and AI models for mental health analysis in social media posts, specifically focusing on detecting low self-esteem in Reddit users’ posts. The researchers propose a higher-level task of reliability analysis, which involves examining the attention mechanism in pre-trained language models for a domain-specific psychology-grounded task. The paper highlights the need to focus more on ‘LoST indicators’ (textual cues suggesting low self-esteem) rather than triggers or consequences of mental disturbance in the posts.
Publication date: 15 Jan 2024
Project Page: Not provided
Paper: https://arxiv.org/pdf/2401.06709
