Research Article
Text Classification to Automatically Identify Online Patients Vulnerable to Depression
470 downloads
@INPROCEEDINGS{10.1007/978-3-319-11564-1_13, author={Taridzo Chomutare}, title={Text Classification to Automatically Identify Online Patients Vulnerable to Depression}, proceedings={Pervasive Computing Paradigms for Mental Health. 4th International Symposium, MindCare 2014, Tokyo, Japan, May 8-9, 2014, Revised Selected Papers}, proceedings_a={MINDCARE}, year={2014}, month={12}, keywords={Online communities Text classification Mood disorders}, doi={10.1007/978-3-319-11564-1_13} }
- Taridzo Chomutare
Year: 2014
Text Classification to Automatically Identify Online Patients Vulnerable to Depression
MINDCARE
Springer
DOI: 10.1007/978-3-319-11564-1_13
Abstract
Online communities are emerging as important sources of support for people with chronic illnesses such as diabetes and obesity, both of which have been associated with depression. The goal of this study was to assess the performance of text classification in identifying at-risk patients. We manually created a corpus of chat messages based on the ICD-10 depression diagnostic criteria, and trained multiple classifiers on the corpus. After selecting informative features and significant bigrams, a precision of 0.92, recall of 0.88, f-score of 0.92 was reached. Current findings demonstrate the feasibility of automatically identifying patients at risk of developing severe depression in online communities.
Copyright © 2014–2024 ICST