There is still much uncertainty about the current COVID-19 outbreak. Modelers are trying to anticipate the future of this pandemic based on relevant parameters driving its evolution. To this aim, the current scientific literature is a core source of information. However, in a context of exponentially increasing numbers of publications, an exhaustive manual analysis remains out of reach.
In this paper, Milliman consultants illustrate the potential and the challenges of using Bidirectional Encoder Representations from Transformers (BERT), a Natural Language Processing (NLP) framework, to automate the task of gathering input information and assisting experts for COVID-19 studies.