Depression Detection Using Sentiment Analysis of Tweets

Abstract

Depression is among the most common mental issue and a major reason for suicides. The rise in the use of social media platforms has given an opportunity for early diagnosis of depression using the everyday language of the users. The motive of this study is to classify tweets having depressive traits among the tweets scraped from Twitter. The present systems fail to capture all the language features used to train the classifiers. This model attempts to maximize the utilization of all the available linguistic features present in the tweets and makes use of proper cleaning and pre-processing techniques for more accurate diagnosis of depression. Rule based Vader Analyzer and a hybrid model of CNN-LSTM is used. The results are calculated and shown using the performance metrics such as recall, precision, F1-score and accuracy.