Sentiment Analysis for Detection of Depressive Users on Social Networks

Depression affects approximately 121 million people worldwide and has as significant an influence on patients as it does their carers. Until recently, psychologists used paper based psychological batteries for screening. New developments in Natural Language Processing however have enabled real-time analysis, offering increases in scalability and accessibility. We built on current knowledge by developing and training several Centralised Neural Network models around a pre-labelled Reddit dataset, comparing post-length and the number of words used per post, selecting the model with the highest F1 score. Our results indicated that although some models performed better than others, neither of our metrics significantly influenced model performance, with changes explained by another factor. Our research was able to yield support for it's use within online forums at a very low cost; justifying further exploration into the use of AI tools for the screening of depression and other mood disorders. Due to a lack of publicly available natural language processing models, our study also resulted in the release of a model for further research by others.