A Novel Approach for Identifying Social Media Posts Indicative of Depression

In this paper, we put forward a novel approach for identifying the social media posts that are indicative of depression with the help of Long Short-Term Memory (LSTM) and Natural Language Processing (NLP) methodologies, Word embedding, n-gram tokenization, and word2vec to identify the text that expresses feelings of depression and its related sentiments. The approach accurately predicts sentiment in the text through Deep Learning, which removes false positives by considering the immediate context of words. The data for this experiment has been scraped from public forums on Reddit-a popular social media website. Labeling is made before analyzing the data, which allows post about a common topic to be grouped. Posts from multiple groups discussing depression and self-harm are taken as the positive class, while posts from miscellaneous, random groups are taken as the negative class. Given the diversity of the negative class, the dataset may be said to be representative of a real-world scenario. The model developed has applications across a broad spectrum of domains, such as deployed on social media and communication forums frequented by children to detect possible harmful tendencies.