Health Text Mining

My research in this area aims to explore health-related issues through social media text mining, with a focus on mental health topics such as depression and suicidal tendencies, often in connection with substance use. By analyzing linguistic and emotional cues and identifying key factors and their interrelationships, this project seeks to enhance mental health monitoring and intervention strategies, particularly for suicide prevention.

Biomedical entity recognition is crucial for advancing research in biomedical and health sciences, enabling tasks like entity linking, relation extraction, and entity resolution. However, low-resource languages like Bangla lack annotated datasets for automatic biomedical entity recognition, limiting progress in applying machine learning models. To address this, BanglaBioMed is introduced as the first Bangla biomedical named entity dataset in IOB format. This dataset serves as a valuable resource for developing biomedical language processing tools for Bangla, annotated with key entity types such as Anatomy, Chemicals and Drugs, Disease and Symptom, and Medical Procedures.

Categorizing communication modes in health-related social media, such as forums and blogs, is essential for understanding health messages and personal experiences. This area remains underexplored for low-resource languages like Bengali. BanglaSocialHealth addresses this gap by introducing the first sentence-level annotated corpus of Bengali health-related posts, classified into four modes: narrative (NAR), informative (INF), suggestive (SUG), and inquiring (INQ). Leveraging machine learning classifiers and transformer models, this work advances Bengali NLP in health research, with applications in information retrieval, question-answering, and public health analysis.