Deep Learning for Suicide Risk Detection in Reddit Posts
摘要
There is an influx of individuals using online platforms to express their feelings of distress when they feel they have no other outlet or support system. Common online warning signs include explicit mentions of suicidal thoughts, writing about death or suicide, drastic changes in behavior or mood, and sharing content related to self-harm. In this paper, we present a natural language processing approach to contribute to the identification of suicidal tendencies in texts written on online forums. We used deep learning approaches such as LSTM, Bi-LSTM, RNN, and GRU, and ensemble learning such as XGBoost, and Random Forest in a Reddit dataset. The model can analyze texts and categorize them according to whether they are suicidal or not. GRU was shown to be the most effective model with an accuracy of 93.25%.