Douglas Teodoro

The paper introduces BioWiC, a benchmark to evaluate how well language models represent biomedical concepts in context, and establishes baselines using various large language models.

This paper uses an ensemble of deep learning language models to classify and rank COVID-19 publications, outperforming standalone models in accuracy and confidence.

The paper presents an information retrieval methodology for finding relevant COVID-19 publications in an infodemic, achieving competitive results in the TREC-COVID challenge.

UPCLASS uses a CNN to classify UniProtKB publications into categories based on protein annotations, achieving higher accuracy than logistic regression and SVM.