Leveraging Deep Learning and Transformer-based Architectures for the Early Identification of Major Depressive Disorder via Longitudinal Linguistic Analysis

Major Depressive Disorder (MDD) represents a significant public health challenge, frequently marked by delays in clinical diagnosis and treatment. The digital footprint generated through social media platforms offers a unique, non-invasive avenue for monitoring psychological states in real-time. This paper proposes a novel framework utilizing a hybrid architecture of Bidirectional Long Short-Term Memory (Bi-LSTM) networks and Transformer models to identify linguistic markers associated with MDD in longitudinal text data. By analyzing shifts in syntactic complexity, sentiment valence, and self-referential language patterns, the proposed method provides a mechanism for early psychological screening. Our experimental results indicate that this model significantly outperforms traditional machine learning benchmarks in sensitivity and specificity. The study concludes with a discussion on the ethical implications of automated mental health monitoring, data privacy, and the necessity of integrating these systems into existing clinical workflows to assist, rather than replace, human psychiatric care.

کلیدواژه ها:

Major Depressive Disorder ، Natural Language Processing ، Deep Learning ، Transformer Architecture ، Mental Health Screening ، Digital Phenotyping ، Affective