Assessing Suicidal Tendencies on Twitter Using BERTicle

Mental health has become a very important health concern nowadays, and peer groups play a vital role in shaping one’s mental growth and health. With the rise of social media, these peer groups can now have a considerably greater amount of influence in your life than they did a decade earlier. Social media has enabled people to post about their thoughts more often and publicly than before. These chains of “posts” have been widely studied to detect extreme emotions like suicidal tendencies in an individual. In addition, the pandemic is causing individuals to experience more stress and fatigue. Social media is all a person has these days to express themselves to others. We have introduced a novel automated framework for detecting and preventing suicidal tendencies through Twitter’s data. This framework combines a BERT [1] model, Twitter’s V2 API, and a Node.js-based Twitter interaction circle algorithm in a single pipeline. The model is trained on more than 9,000 tweets with a prediction accuracy of 94.7%. The interaction circles are generated using the classical distance formula, and the Twitter API V2 facilitates all the Twitter interactions. We propose a new framework in this article to predict and prevent suicidal tendencies on Twitter.