Binary classification of English Reddit posts self-reporting a social anxiety disorder diagnosis
Singh, S.; Bedi, J.
Show abstract
This paper presents the system developed by Team ThaparUni for the Social Media Mining for Health Applications (SMM4H) 2023 Shared Task 4. The task involved binary classification of English Reddit posts, focusing on self-reporting social anxiety disorder (SAD) diagnoses. The final system employed a combination of three models: RoBERTa, ERNIE, and XLNet, and results obtained from all three models were integrated. The results, specifically in the context of mental health-related content analysis on social media platforms, show the possibility and viability of using multiple models in binary classification tasks.
Matching journals
The top 4 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Machine Learning Models Predict the Emergence of Depression in Argentinean College Students during Periods of COVID-19 Quarantine 92%
- Passive sensing data predicts stress in university students: A supervised machine learning method for digital phenotyping 92%
- Understanding Psychiatric Illness Through Natural Language Processing (UNDERPIN): Rationale, Design, and Methodology 91%
Similar papers in this journal
- Users’ Reactions on Announced Vaccines against COVID-19 Before Marketing in France: Analysis of Twitter posts 94%
- Developing an automatic system for classifying chatter about health services from Twitter: A case study for Medicaid 93%
- Information retrieval in an infodemic: the case of COVID-19 publications 91%
Similar papers in this journal
- Listening to mental health crisis needs at scale: using Natural Language Processing to understand and evaluate a mental health crisis text messaging service 96%
- Development and Validation of a Machine Learning Model Integrated with the Clinical Workflow for Inpatient Discharge Date Prediction 90%
- CharMark: A Markov Approach to Linguistic Biomarkers in Dementia 89%
Similar papers in this journal
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.