A BERT-Based Model for Detecting Depression in Diabetes-Related Social Media Posts

Rdouan Faizi, Bouchaib Bounabat, Mahmoud El Hamlaoui

2025

Abstract

This paper introduces a BERT-based model for detecting depression in diabetic social media posts. Based on transformer-based language models, the proposed approach is specifically designed to capture the specific linguistic patterns that are indicative of depressive symptoms. The model was trained on a dataset of comments retrieved from diabetes-related YouTube channels, which were then manually annotated as either ‘Depression’ or ‘Well-being’. Through extensive experimentation, the model achieved a high classification accuracy of 93% on the test set. These findings highlight its potential as an effective tool for automated mental health monitoring in at-risk populations, particularly those coping with chronic health conditions such as diabetes.

Paper Citation

in Harvard Style

Faizi R., Bounabat B. and El Hamlaoui M. (2025). A BERT-Based Model for Detecting Depression in Diabetes-Related Social Media Posts. In Proceedings of the 14th International Conference on Data Science, Technology and Applications - Volume 1: DATA; ISBN 978-989-758-758-0, SciTePress, pages 627-632. DOI: 10.5220/0013636100003967

in Bibtex Style

@conference{data25,
author={Rdouan Faizi and Bouchaib Bounabat and Mahmoud El Hamlaoui},
title={A BERT-Based Model for Detecting Depression in Diabetes-Related Social Media Posts},
booktitle={Proceedings of the 14th International Conference on Data Science, Technology and Applications - Volume 1: DATA},
year={2025},
pages={627-632},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013636100003967},
isbn={978-989-758-758-0},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 14th International Conference on Data Science, Technology and Applications - Volume 1: DATA
TI - A BERT-Based Model for Detecting Depression in Diabetes-Related Social Media Posts
SN - 978-989-758-758-0
AU - Faizi R.
AU - Bounabat B.
AU - El Hamlaoui M.
PY - 2025
SP - 627
EP - 632
DO - 10.5220/0013636100003967
PB - SciTePress