Common N-Gram Method: A Promising Approach to Detecting Mental Health Disorders on Social Media
Date
2023-04-13
Authors
Agarwal, Harshit
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This paper addresses the mental health challenges posed by the COVID-19 pandemic and the lack of reliable and accessible diagnostic tools for mental health conditions. The dataset used in this research consists of over 2 million posts from the social media platform called Reddit. 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 Common N-gram (CNG) method that offers comparable results to the state-of-the-art CNN-LSTM model and is less resource-intensive. The CNG method shows better performance in comparison to the CNN-LSTM model and SVM, baseline model, in multi-classification 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 common N-gram technique.
Description
Keywords
Natural Language Processing, Mental Health, Common N-gram Method, Word Embeddings, Neural Networks, Deep Learning, Machine Learning