Open AccessCovidBERT-Biomedical Relation Extraction for Covid-19
Author(s) -
Shashank Hebbar,
Ying Xie
Publication year - 2021
Publication title -
proceedings of the ... international florida artificial intelligence research society conference
Language(s) - English
Resource type - Journals
eISSN - 2334-0762
pISSN - 2334-0754
DOI - 10.32473/flairs.v34i1.128488
Subject(s) - covid-19 , computer science , transformer , explosive material , clinical trial , relationship extraction , pandemic , data science , disease , infectious disease (medical specialty) , relation (database) , medicine , data mining , engineering , history , virology , pathology , voltage , outbreak , electrical engineering , archaeology
Given the ongoing pandemic of Covid-19 which has had a devastating impact on society and the economy, and the explosive growth of biomedical literature, there has been a growing need to find suitable medical treatments and therapeutics in a short period of time. Developing new treatments and therapeutics can be expensive and a time consuming process. It can be practical to re-purpose existing approved drugs and put them in clinical trial. Hence we propose CovidBERT, a biomedical relationship extraction model based on BERT that extracts new relationships between various biomedical entities, namely gene-disease and chemical-disease relationships. We use the transformer architecture to train on Covid-19 related literature and fine-tune it using standard annotated datasets to show improvement in performance from baseline models. This research uses the transformer BERT model as its foundation and extracts relations from newly published biomedical papers.