Staff Profile Pages

General Vs Domain-Specific Language Models for Depression Detection in Social Media and Clinical Texts: A Comparative Study

Authors: Bouchriha, K., Guelzim, I., Hirchoua, B., Nait-Charif, H.

Journal: 2025 IEEE International Conference on Emerging Trends in Engineering and Computing Etecom 2025

Publication Date: 01/01/2025

DOI: 10.1109/ETECOM66111.2025.11319047

Abstract:

Depression is one of the most serious medical problems. The World Health Organization (WHO) estimates that depression affects more than 264 million people worldwide, in all age groups. Depression is a major contributing factor to suicide. Alarming the urgent need for effective methods to detect and treat depression.In this paper, we conduct a comparative study of different situations applying general and specific domain language models. Specifically, we consider, among others, BERT, BioBERT, and ClinicalBERT using social media texts and medical notes as data sources. The results show that the models work better when the data sources are close to their training domain, but performance decreases when they are older or come from a different context. These findings highlight the importance of updating and adapting models in the field of mental health.

Source: Scopus