Applying BERT Model for Early Detection of Mental Disorders Based on Text Input

In today's digital era, awareness of mental health issues is growing significantly. Many individuals are now more open about sharing their psychological conditions through written texts on social media, forums, and surveys. This phenomenon presents an opportunity to leverage technology for the automatic detection of mental disorders through text analysis. This study aims to implement the Bidirectional Encoder Representations from Transformers(BERT) model to identify mental health conditions such as depression, bipolar disorder, anxiety, suicidal tendencies, and others. The dataset was sourced from Kaggle and underwent several preprocessing stages, including data cleaning, tokenization, and text classification model training. This BERT model achieved strong performance, with an accuracy of 91% and an average F1-Score of 0.91. These results demonstrate the model's effectiveness in identifying various psychological expressions. The findings highlight the potential for developing early detection systems that are faster, more objective, and widely accessible. However, this study acknowledges limitations in dataset diversity, suggesting future work to incorporate more varied data sources and explore other NLP models to enhance detection accuracy and coverage.