Forensic suicidal inquiry of depressed individuals using LSTM and convolutional neural networks

Abstract

Social media sites such as Twitter, Facebook, Tumblretc, are vastly popular among the general population. People post updates, tweets etc., and almost 75% of the times, these posts are a combination of emotions. The idea is to analyze suicidal-depression tendencies in adults with traumatizing experiences or socio-economic difficulties. This makes the overall analysis of sentiments especially extremely complex, which we aim to resolve here in this project by breaking down all the sentences into individual words, and along with emoticons and hashtags, converting each one of them into tokens, and then applying deep learning algorithms on the same, to accurately determine the sentiments of given messages. The objective of the project undertaken is to determine the suicidal-sentiment of various depressed individuals, and how likely is it that they are inclined to commit suicide on the basis of their tweets.