GT CSE 6250 Spring Project

Our team will attempt to validate the results of this paper and show that a CNN deep learning model can better categorize social media posts according to risk for suicide comparted to traditional ML models.

Presentation: https://www.youtube.com/watch?v=q8swAqHIBTw&lc=Ugy0EMoHcx6Ly5biuy94AaABAg.

Google Collab: https://colab.research.google.com/drive/1fVu84WMKqF3Fa7VqJ3zFT6rt5D_TMR9e?usp=sharing

Full Source Code Instructions

The full source code includes procedures for re-building a dataset, running and optimizing the models.

(Note that python 3.11 was used to build this project.)

1. Create python environment and install required modules (ie. pip install -r requirements.txt).
2. Open and run all cells in the file CSE-6250-Team-A1-Full Source Code.ipynb

Resources

Paper: Knowledge-Aware Assessment of Severity of Suicide Risk for Early Intervention

Authors: Manas Gaur, Amanuel Alambo, Joy Sain, Ugur Kursuncu, Krishnaprasad Thirunarayan, Ramakanth Kavuluru, Amit Sheth, Randy Welton, Jyotishman Pathak

Source Code: Github

Dataset: Github

Additional Resources

1. Leveraging Reddit for Suicidal Ideation Detection: excellent article that reviews over 100 studies related to this topic.
2. Suicide Risk Assessment Reference Guide: short guide for diagnosing suicidal behavior (manually)