Suicide Ideation Detection in Social Media Forums

Numerous initiatives have been developed to prevent suicide. Still, people are not seeking help considering it will dishonor them in society, and they are unwilling to reveal their identity. However, thanks to social media platforms, people are more willing to express themselves anonymously as they do not directly face anyone. Our work aims to use this vast data to identify texts containing suicidal thoughts and help early detection of suicide using machine learning and deep learning techniques. The dataset used is collected by taking posts from "SuicideWatch" and "teenagers" Subreddits of Reddit using, "Pushshift" API. The proposed model uses a Universal Sentence Encoder to encode the text and a Fully connected Neural Network(FCNN) to classify text into suicide, and non-suicide, which achieved a better accuracy compared to the existing methods.