Detecting Suicidal Intentions through Analysis of Social Media Content

Suicide prevention remains a critical challenge, and the detection of suicidal intentions through social media content has emerged as a promising method to identify individuals at risk. This study investigates the use of advanced machine learning, deep learning, and transformer-based models to detect signs of suicidal ideation in bilingual social media posts. A carefully curated dataset, consisting of Bengali and English posts which were collected from various social media like Facebook, Twitter and Youtube, was annotated with expert verification to ensure accuracy. The models were evaluated on their ability to capture the complex linguistic and cultural nuances inherent in both languages. Among the models tested, MultilingualBERT achieved the best performance with a F1 score of 0.90, demonstrating its effectiveness in handling the intricacies of bilingual data. The results emphasize the effectiveness of advanced models in practical applications for mental health monitoring and suicide prevention, offering a powerful tool for early detection and intervention.