This paper evaluates the performance of “AAST-NLP” in the Social Media Mining for Health (SMM4H) Shared Tasks 3 and 6, where more than 20 teams participated in each. We leveraged state-of-the-art transformer-based models, including Mistral, to achieve our results. Our models consistently outperformed both the mean and median scores across the tasks. Specifically, an F1-score of 0.636 was achieved in classifying the impact of outdoor spaces on social anxiety symptoms, while an F1-score of 0.946 was recorded for the classification of self-reported exact ages.

This article describes the participation of the UU_TAILS team in the 2019 MEDIQA challenge intended to improve domain-specific models in medical and clinical NLP. The challenge consists of 3 tasks: medical language inference (NLI), recognizing textual entailment (RQE) and question answering (QA). Our team participated in tasks 1 and 2 and our best runs achieved a performance accuracy of 0.852 and 0.584 respectively for the test sets. The models proposed for task 1 relied on BERT embeddings and different ensemble techniques. For the RQE task, we trained a traditional multilayer perceptron network based on embeddings generated by the universal sentence encoder.