<![CDATA[Online platforms contain user-generated text that may signal depression, suicidal ideation, and suicide risk. This Systematic Literature Review (SLR) examines machine learning (ML), deep learning (DL), and natural language processing (NLP) approaches to text-based mental health detection. Following PRISMA 2020, studies published from 2020 to 2025 were retrieved from ScienceDirect and screened using criteria. Twenty-five studies were included. The studies analyzed text from Twitter/X and Reddit, applying traditional ML classifiers, word embeddings, recurrent neural networks, attention mechanisms, ensemble methods, and transformer models. Performance was evaluated using accuracy, precision, recall, F1-score, and AUC. However, variations in datasets, classification targets, and evaluation procedures limit comparisons. The review identifies challenges including data scarcity, class imbalance, annotation subjectivity, contextual ambiguity, generalization, interpretability, and privacy. The findings emphasize the need for representative datasets, validation, explainable models, and data governance. These approaches may support the identification of mental health signals but should not replace professional assessment.]]>
Copyrights © 2026
