Sentiment Analysis for Mental Health

Abstract

In today's world, mental health is a growing problem and people need better ways to understand and deal with it. That is why tools that can automatically analyse how someone is feeling have become more important. Sentiment analysis, a subset of natural language processing (NLP), offers a powerful approach to understanding emotions and psychological states expressed in text. In this study, we present a Logistic Regression-based model trained to predict the nature of individuals’ mental health based on textual data. The dataset, sourced from Kaggle, consists of labeled mental health-related texts reflecting a range of emotions, psychological states, and conditions. Before training the model, we cleaned the text data through steps like normalizing words, splitting them into tokens, and converting them into numbers using TF-IDF. The model is trained to look at a piece of text and decide which mental health condition it is most likely linked to, or whether the person seems fine. We tested the model using accuracy, precision, recall, and F1-score, and the results showed it can predict mental health conditions reasonably well. Our results show that sentiment analysis can work alongside traditional methods of mental health diagnosis by giving a fast, automated way to understand how people feel based on what they write. This work shows that machine learning has real potential in the mental health space, and future research can build on this to develop stronger models that could be used for early diagnosis and support.