Shayona@SMM4H23: COVID-19 Self diagnosis classification using BERT and LightGBM models
This work addresses automated detection of health-related self-reports in social media data, but it is incremental as it combines existing models without introducing new methods.
The paper tackled binary classification of English tweets and Reddit posts for self-reported COVID-19 and social anxiety disorder diagnoses, achieving the highest F1-score of 0.94 in the COVID-19 task.
This paper describes approaches and results for shared Task 1 and 4 of SMMH4-23 by Team Shayona. Shared Task-1 was binary classification of english tweets self-reporting a COVID-19 diagnosis, and Shared Task-4 was Binary classification of English Reddit posts self-reporting a social anxiety disorder diagnosis. Our team has achieved the highest f1-score 0.94 in Task-1 among all participants. We have leveraged the Transformer model (BERT) in combination with the LightGBM model for both tasks.