Prediction of Depression using Emotion and Sentiment Analysis
Anurag Pal, Tanmoy Maitra, Debasis Giri · 2023
In today’s world, people are facing a lot of mental health issue. Among which, depression is the most prevalent ones. Many of the times, it can be detected by a person’s on-line behaviour like what type of comments/posts he posts on-line. So social media companies can use this information to uplift the mood of the depressed user by showing positive/joyful contents. That is why analyzing depression through texts is so important. In this paper, we have trained and built an emotion classifier model using a different dataset. Then, we have fitted this model onto another dataset and found out the emotions of text. We have also analyzed the sentiments separately. Thus, generating two new columns for emotion and sentiment respectively. Then we group by the emotions, sentiment and the column that shows whether the text is depressed or not. We find some relations between these three and this helps us to build a depression predictor function. As per the results, we get an average accuracy with traditional classifiers like, Naive Bayes. We can directly predict depression just by looking at the emotion and sentiment of a user. Also, this method can be used as a supportive method for the better models to optimize them even more. Besides all these, we had some interesting findings. We have found out that which combinations of emotion and sentiment are more prone to depression than others, and which combination is just at the edge of depression.