A Multilevel Depression Detection from Twitter using Fine-Tuned RoBERTa

Depression stands out as the most commonly occurring psychiatric disorder worldwide. Extreme levels of depression can make people take drastic steps. However, early-stage depression detection and treatment can save many lives. Different social media platforms play a vital role in analyzing depression in its early stages. Hence, this work aims to identify and intensify depression through social media posts. This research highlights the potential of social media for the early detection of depression and its significance in aiding psychiatrists to prevent suicide by identifying depression in its initial phases. The study employs a fine-tuned RoBERTa model to forecast multilevel depression (mild, moderate, and severe) from Twitter posts. The research aim to enhance comprehension of depression severity and its impact on mental health by conducting a thorough context analysis utilizing a fine-tuned RoBERTa model. The model proposed in this study exhibits a 90% of accuracy in predicting multilevel depression. The analysis illuminates the effectiveness of the suggested framework and its possible implementation in mental health research and treatment.