Suicidal Behaviour Detection and Prevention

Abstract

The global use of social media has reshaped human relationship with liberty to exchange experience and emotions without limitation. Free communication with the risk of suicidal content, prevalent among the youth generation, requires efficient monitoring and intervention. The project currently utilizes the BERT-based deep learning models to detect and prevent suicide risk based on text analysis. The system assigns the content either as \"Sad\" or \"Not Sad\" depending on whether social media postings are being monitored. If the post is categorized to \"Sad,\" the system sends motivational quotes through SMS and Telegram through Twilio and Telethon following an auto-response setting. The system also forwards all distress posts to a suicide help line for immediate expert intervention. Successful Natural Language Processing (NLP) and artificial communication integration ensures successful suicide risk assessment for successful suicide prevention against catastrophic consequences. Suicidal prevention theory of prevention allows for reinforcement of online tracking of mental state and hence is a vital emotional well-being and crisis therapy.

Introduction

The paper presents an AI-based suicide prevention system that utilizes Natural Language Processing (NLP) and deep learning, particularly BERT and RoBERTa models, to detect suicidal behavior through social media posts. It aims to overcome the limitations of manual monitoring by providing real-time, scalable, and automated detection of emotional distress.

Key Features of the System:

• Sentiment Classification:
  The system uses transformer-based models like BERT and RoBERTa to classify social media posts as "Sad" or "Not Sad", allowing early identification of users showing signs of emotional distress.

• Real-Time Intervention:
  If a post is flagged as "Sad", the system sends automated care messages via SMS (Twilio) and Telegram (Telethon) to the user, offering emotional support. Simultaneously, it alerts suicide prevention hotlines for further professional assistance.

• Architecture and Process:
  The system consists of:

  1. Data Collection & Preprocessing: Tokenization, stopword removal, and vectorization.

  2. Sentiment Analysis: Using multiple NLP models for robust classification.

  3. Automated Response: Messaging and hotline alerts upon distress detection.

• Data Augmentation Techniques:
  To improve model generalization, techniques like synonym replacement, back translation, and sentence shuffling are used, helping the system recognize a wide range of linguistic expressions of distress.

• Performance Metrics:
  The models are evaluated using Precision, Recall, and Mean Average Precision (mAP):

  • RoBERTa shows the best overall performance with:

    • Precision: 0.99

    • Recall: 0.89

    • mAP: 0.94

• Literature Insights:
  The review includes recent studies employing BERT variants, CNNs, and multimodal approaches. These highlight the effectiveness of transformer-based models in capturing nuanced emotional content and emphasize the growing role of AI in mental health applications.

Conclusion

Suicidal behavior detection plays a crucial role in mental health intervention, where timely identification can significantly improve support and prevention efforts. Deep learning-based Natural Language Processing (NLP) models, particularly transformer-based architectures, have demonstrated their effectiveness in analyzing social media posts for distress signals. This study explores the application ofRoBERTa, to develop an efficient suicidal behavior detection system. A curated dataset of social media posts labeled with emotional sentiments is used to train and evaluate these models, ensuring robust learning and accurate predictions. The dataset undergoes preprocessing techniques such as tokenization, stop-word removal, and augmentation methods like synonym replacement and back translation to enhance detection performance. Each NLP model is assessed for its efficiency, classification accuracy, and reliability in detecting suicidal tendencies. The results indicate that advanced transformer architectures significantly improve detection accuracy, making them suitable for real-time mental health monitoring. Performance metrics such as precision, recall, and mean Average Precision (mAP) were used for evaluation, with RoBERTa achieving the highest precision. By automating the detection process, this system reduces dependency on manual assessment, minimizing oversight and assisting mental health professionals in early intervention. The study highlights the potential of deep learning in revolutionizing mental health support, offering a scalable and effective solution for suicide prevention.

References

[1] Parsapoor, M., Koudys, J. W., & Ruocco, A. C. (2023). Suicide risk detection using artificial intelligence: The promise of creating a benchmark dataset for research on the detection of suicide risk. Frontiers in Psychiatry, 14, 1186569. [2] Tang, H., Rekavandi, A. M., Rooprai, D., Dwivedi, G., Sanfilippo, F., Boussaid, F., &Bennamoun, M. (2023). Analysis and evaluation of explainable artificial intelligence on suicide risk assessment. Scientific Reports, 13(1), 1-12. [3] Ormerod, C. M., Patel, M., & Wang, H. (2023). Using language models to detect alarming student responses. arXiv preprint arXiv:2305.07709. [4] Elsayed, N., ElSayed, Z., & Ozer, M. (2024). CautionSuicide: A deep learning-based approach for detecting suicidal ideation in real-time chatbot conversations. arXiv preprint arXiv:2401.01023. [5] Badian, Y., Ophir, Y., Tikochinski, R., Calderon, N., Klomek, A. B., & Reichart, R. (2023). A picture may be worth a thousand lives: An interpretable artificial intelligence strategy for predictions of suicide risk from social media images. arXiv preprint arXiv:2302.09488. [6] Chen, Y., Li, J., Song, C., Zhao, Q., Tong, Y., & Fu, G. (2024). Deep Learning and Large Language Models for Audio and Text Analysis in Predicting Suicidal Acts in Chinese Psychological Support Hotlines. arXiv preprint arXiv:2409.06164. [7] Ananya G., Bhumika J., Minakhi R.(2023). Prediction of Suicide Rates in India Using Machine Learning. 2023 IEEE International Conference on Advances in Science and Technology (ICAST), 1-6. [8] Maini, M., Srivastava, P., Soni, H., Hemprasanna, & Pillai, A. S.(2023). Prediction of Suicide Rates in India Using Machine Learning. Proceedings of the 2023 IEEE International Conference on Advances in Science and Technology (ICAST), 1-6. [9] Wibhowo, C., & Sanjaya, R. (2021). Virtual Assistant to Suicide Prevention in Individuals with Borderline Personality Disorder. Proceedings of the 2021 International Conference on Control, Automation, and Information Sciences (ICCAIS). [10] Qiu, H., Ma, L., & Lan, Z. (2024). PsyGUARD: An automated system for suicide detection and risk assessment in psychological counseling. arXiv preprint arXiv:2409.20243. [11] Tank, C., Mehta, S., Pol, S., Katoch, V., Anand, A., Jaiswal, R., & Shah, R. R. (2024). Su-RoBERTa: A semi-supervised approach to predicting suicide risk through social media using base language models. arXiv preprint arXiv:2412.01353. [12] Liu, S., Lu, C., Alghowinem, S., Gotoh, L., Breazeal, C., & Park, H. W. (2022). Explainable AI for suicide risk assessment using eye activities and head gestures. arXiv preprint arXiv:2206.07522. [13] Liu, M., & Liu, S. (2021). Stability Analysis of a Stochastic SIRS Epidemic Model with a General Nonlinear Incidence Rate. Proceedings of the 2021 IEEE International Conference on ArtificialIntelligence and Computer Applications (ICAICA), 945-949. [14] Cui, Z., Lei, C., Wu, W., Duan, Y., Qu, D., Wu, J., Chen, R., & Zhang, C. (2024). Spontaneous speech-based suicide risk detection using Whisper and large language models. arXiv preprint arXiv:2406.03882. [15] Nfissi, A., & Ouni, A. (2024). Unlocking the emotional states of high-risk suicide callers through speech analysis. Proceedings of the 18th IEEE International Conference on Semantic Computing (ICSC). [16] Nock, M. K., Borges, G., Bromet, E. J., Cha, C. B., Kessler, R. C., & Lee, S. (2008). Suicide and suicidal behavior. Epidemiologic Reviews, 30(1), 133–154. [17] Atmakuru, A., Shahini, A., Chakraborty, S., Seoni, S., Salvi, M., Hafeez-Baig, A., Rashid, S., Tan, R. S., Barua, P. D., Molinari, F., & Acharya, U. R. (2024). Artificial intelligence-based suicide prevention and prediction: A systematic review (2019–2023). Biomedical Signal Processing and Control, 85, 104965. [18] Ji, S., Pan, S., Li, X., Cambria, E., Long, G., & Huang, Z. (2020). Suicidal ideation detection: A review of machine learning methods and applications. IEEE Transactions on Computational Social Systems, 7(3), 1145-1158. [19] Singh, P. K., & Kumar, A. (2023). Designing a model for suicidal behaviour detection using machine learning. International Journal of Advanced Research in Computer Science, 14(3), 123-130. [20] Levkovich, I., & Omar, M. (2024). Evaluating BERT-based and large language models for suicide detection, prevention, and risk assessment: A systematic review. Journal of Medical Systems, 48(1), 113.

Copyright

Copyright © 2025 Bhargavi Sankarana, Jai Bhavani Muppidi, Yamini Appikatla, Dr. G. Krishna Kishore. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.