Text classification models for personality disorders identification

This research focuses on identifying personality disorders in individuals using their social media text. We developed a unique collection of words (PD-Corpus) and a dataset (PD-TXT), which includes texts marked with different personality disorder traits. Our goal was to classify these texts into six types of personality disorders, using Natural Language Processing (NLP) classification models. The results showed that our transformer-based models, especially the BERT-base-uncased model, were more effective than traditional methods, achieving a 74.7% success rate in correctly classifying these disorders. Also, our models consistently outperform existing literature baseline models on the PD-TXT dataset, showcasing significant enhancements. This study presents a new way to predict personality disorders through linguistic analysis and highlights the potential for further research combining language studies with mental health.