Common N-Gram Method (CNG): A Promising Approach to Detecting Mental Health Disorders on Social Media

Technology and social media's growing prevalence, especially during the COVID-19 pandemic, has contributed to a rise in mental health issues. This research discusses novel applications of natural language processing which can help develop more effective and accessible diagnostic tools for mental health illnesses. To enhance the realism of our model, we created a biased dataset that reflects the real-world ratios of mental illness prevalence. The proposed solution is the Common N-grams (CNG) method that offers comparable results to the state-of-the-art CNN-LSTM model and is less resource-intensive. The CNG method performs better than the CNN-LSTM model and Support Vector Machine (SVM), baseline model, in multiclassification tasks. The CNN-LSTM surpasses performance in binary tasks compared to the best score reported in the previous study with the same dataset. The study also highlights the usefulness of the Relative N-Gram Signature method to analyze the classification decision of the CNG technique. The proposed solutions offer practical and accessible options for individuals seeking reliable and accurate mental health support.