The TREC Health Misinformation track focuses on discerning reliable from unreliable information and correct from incorrect information. This problem is very common in Web Search results and it is especially critical when it is related to health content. This year's task focuses on COVID-19 and SARS-CoV-2 misinformation. In our experiments, we applied a BM25 retrieval baseline as a first step. Afterwards, we used a reliability classifier recently developed by our team. Finally, we also experimented with BERT-based variants that attempt to estimate similarity between sentences.
Keywords: misinformation, COVID-19, reliability