This paper presents our models for the Social Media Mining for Health 2024 shared task, specifically Task 5, which involves classifying tweets reporting a child with childhood disorders (annotated as “1”) versus those merely mentioning a disorder (annotated as “0”). We utilized a classification model enhanced with diverse textual and language model-based augmentations. To ensure quality, we used semantic similarity, perplexity, and lexical diversity as evaluation metrics. Combining supervised contrastive learning and cross-entropy-based learning, our best model, incorporating R-drop and various LM generation-based augmentations, achieved an impressive F1 score of 0.9230 on the test set, surpassing the task mean and median scores.

This paper describes models developed for the Social Media Mining for Health (SMM4H) 2022 shared tasks. Our team participated in the first subtask that classifies tweets with Adverse Drug Effect (ADE) mentions. Our best-performing model comprises of a template augmented task adaptive pre-training and further fine-tuning on target task data. Augmentation with random prompt templates increases the amount of task-specific data to generalize the LM to the target task domain. We explore 2 pre-training strategies: Masked language modeling (MLM) and Simple contrastive pre-training (SimSCE) and the impact of adding template augmentations with these pre-training strategies. Our system achieves an F1 score of 0.433 on the test set without using supplementary resources and medical dictionaries.