Identification of depression and PTSD among Twitter users using pre-trained language model

ACL ARR 2025 February Submission1218 Authors

Abstract: Suicide is a global health issue and early diagnosis is necessary for effective treatment. Recent advancements in natural language processing has aided the identification of mental health disorders in social media. This paper investigated the efficacy of pre-trained language model (PLM) in identifying depression and post-traumatic stress disorder (PTSD) with Twitter data. Leveraging the CLPysch 2015 dataset (which constitutes of tweets from users with depression, PTSD and neither condition), we implemented various experimental designs using Long Short Term Memory (LSTM) and attention. The results demonstrate that while detecting specific mental health issues is still difficult, the detection of general mental health conditions improves with the implementation of attention. The results provide insights into the strengths and weaknesses of these models in identifying mental health issues from social media content, with potential implications for improving mental health monitoring.

Paper Type: Short

Research Area: NLP Applications

Research Area Keywords: NLP Applications, Healthcare applications

Contribution Types: Model analysis & interpretability, Publicly available software and/or pre-trained models

Languages Studied: English

Submission Number: 1218