The Role of Natural Language Processing in Early Detection of Mental Health Conditions from Social Media Data

Abstract

Mental health disorders such as anxiety, depression, and schizophrenia are increasing rapidly and affect a significant proportion of the global population. As a result, the affected patients suffer negative consequences such as high financial costs of treatment and a poor quality of life. The reliance on traditional clinical methods to diagnose mental health problems further leads to delays in identifying the disorders among affected individuals. An emerging approach to address the delay is the adoption of artificial intelligence (AI) through natural language processing (NLP) models, which can evaluate real-time social media content to identify individuals at risk of mental health problems. The current research sought to identify how NLP techniques could be adopted for the early diagnosis and detection of mental health illnesses from social media interactions. Data was collected using the scoping review method, where 20 qualitative peer reviewed journal articles were identified and assessed. To evaluate the findings obtained in the study, thematic analysis was adopted. The generated insights indicated that using deep learning techniques, including recurrent neural networks (RNNs) and classification machine learning methods, such as decision trees (DT), facilitated the detection of mental health illnesses. Further insights revealed that techniques such as data anonymisation were effective for privacy preservation, and explainable AI (XAI) were useful in upholding the privacy of user data during the data collection phase. Additionally, various advantages of NLP models were elaborated, including accuracy, generalisability, and fairness. However, challenges such as risks of bias and breaching the privacy of user data were also identified. In future work, there is a need to investigate how the NLP models can be enhanced further by integrating more technologies, such as big data.