Research Article
Revealing Mental Disorders Through Stylometric Features in Write-Ups
@INPROCEEDINGS{10.1007/978-3-031-34776-4_14, author={Tamanna Haque Nipa and A. B. M. Alim Al Islam}, title={Revealing Mental Disorders Through Stylometric Features in Write-Ups}, proceedings={Mobile and Ubiquitous Systems: Computing, Networking and Services. 19th EAI International Conference, MobiQuitous 2022, Pittsburgh, PA, USA, November 14-17, 2022, Proceedings}, proceedings_a={MOBIQUITOUS}, year={2023}, month={6}, keywords={Stylometric Marker Imbalanced Dataset Personal Pronoun}, doi={10.1007/978-3-031-34776-4_14} }- Tamanna Haque Nipa
A. B. M. Alim Al Islam
Year: 2023
Revealing Mental Disorders Through Stylometric Features in Write-Ups
MOBIQUITOUS
Springer
DOI: 10.1007/978-3-031-34776-4_14
Abstract
Mental disorders present one of the leading causes of worldwide disability and have become a major social concern, as the symptoms behind mental disorders are almost hidden. Most of the conventional approaches used for diagnosing and identifying mental disorders rely on oral conversations (through interviews) having a limited focus on write-ups. Therefore, in this study, we attempt to explore identifying different types of mental disorders among people through their write-ups. To do so, we collect a total of 6893 posts and discussions that appeared in different problem-specific Internet forums and utilize them to identify different types of mental disorders. Leveraging appropriate machine learning algorithms over the collected write-ups, our study can categorize Depression, Schizophrenia, Suicidal Intention, Anxiety, Post Traumatic Stress Disorder (PTSD), Borderline Personality Disorder (BPD), and Eating Disorder (ED). To achieve a balanced dataset in the process of our study, we apply a combined sampling approach and achieve up to 89% accuracy in the identification task. We perform varied exploration tasks in our study covering 5-fold cross-validation, 5-times repetition on the used dataset, etc. We explain our findings in terms of precision, recall, specificity, and Matthews correlation coefficient to demonstrate the capability of our proposed approach in identifying mental disorders based on write-ups.
