Comparative Analysis of NLP Models for Detecting Depression on Twitter

Depression is a serious mental health issue affecting a significant portion of the world’s population. With the widespread use of social media platforms, researchers have explored the possibility of utilizing natural language processing (NLP) techniques to detect signs of depression in users’ posts. In this paper, we present a comparative analysis of six different NLP models, namely BERT, RoBERTa, DistilBERT, ALBERT, Electra, and XLNet, for depression detection on Twitter data. The experiments compare the performance of different models, and the results reveal that the highest-performing models include XLNet, DistilBERT, and RoBERTa with the accuracies over 99%.