Artificial intelligence-based text mining for COVID-19

Project: Research

Research areas and keywords

UKÄ subject classification

  • Medical and Health Sciences
  • Humanities
  • Engineering and Technology

Keywords

  • SARS-CoV-2, COVID-19, BioNLP, natural language processing, text mining, Artificial Intelligence

Description

In the current COVID-19 crisis, authorities, researchers and pharmaceutical companies are rushing to develop new treatments and public health strategies. For this, they need to rapidly review a mountain of scientific reports and other texts. Currently, there are already over 120000 articles (https://pages.semanticscholar.org/coronavirus-research) containing information on coronaviruses and the literature expands at an accelerating rate as teams all across the world scramble to share information. Consequently, reviewing texts manually is no longer possible. In addition, humans are unable to connect all of the fragmented information that is scattered in pieces over such a vast body of literature. Artificial intelligence-based text mining has reached sufficient maturity to solve this problem at least in part by flagging the most relevant texts for expert review and even extracting and summarizing some type of information such as drug-disease or virus-protein relationships.
Our interdisciplinary team, which comprises biomedical experts, bioinformaticians, computational linguists and computer scientists is developing an AI-based COVID-19 text mining toolbox. Our goal is to help researchers, doctors, nurses, public health organizations, pharmaceutical industry and policy makers to make the best evidence-based decisions despite the urgency and overwhelming amount of information. We also want to enable the public and journalists to access scientific findings despite a lack of training in medical literature search strategies.
We have already released a first part of the English version of our toolbox: https://arxiv.org/abs/2003.09865. We are now working hard to improve the toolbox and to add capabilities for analysing Swedish texts. Once the system is completed, we will use it to extract key knowledge needed by other COVID-19 researchers and organizations. For example, we have received a number of key questions from our collaborators at the Coalition for Epidemic Preparedness Innovations, which coordinates many of the COVID-19 vaccine development programs. We are also closely connected to other Swedish COVID-19 research groups as part of the SciLifeLab national COVID-19 program.
StatusActive
Effective start/end date2020/01/20 → …

Participants

Related research output

View all (1)