Salesforce Launches COVID-19 Search to Sift Through Pandemic Studies
COVID-19 is generating research at a rate commensurate with its infections. As millions of cases spread across the world, hundreds of thousands of research papers are pouring into journals, overwhelming researchers in their frantic work to develop vaccines, treatments, and accurate epidemiological models. Now, Salesforce is introducing COVID-19 Search, an AI-powered search engine that aims to help scientists find the most relevant COVID-19 research for their needs.
By May, more than 138,000 research papers had been published on COVID-19 – and more than a million are expected by December. These research papers are being coordinated into the COVID-19 Open Research Dataset (CORD-19) by the White House, the National Institutes of Health (NIH), Georgetown University, and other leading research groups.
COVID-19 Search is Salesforce’s response to the CORD-19 Challenge, which tasked AI experts with creating text and data mining tools to help medical researchers find answers to COVID-19-related questions. The company brought together its experts in natural language processing (NLP), and over a few months, they worked to develop the search engine with a couple of key priorities in mind: it had to interpret the proper meaning of a search, and it should quickly present relevant passages in results.
“COVID-19 Search addresses this by combining text retrieval and NLP — including semantic search, state of the art question answering, and abstractive summarization — to better understand the question and surface the most relevant scientific results,” wrote Andre Esteva (head of medical AI at Salesforce Research) and Anuprit Kale (lead data scientist at Salesforce) in a blog post announcing the tool.
Using NLP, the tool emphasizes semantic search that is able, for instance, to differentiate between a query like “What expression pathways does SARS-CoV-2 induce?” and “What is the expression pathway of SARS-CoV-2?” To resolve these queries, the search engine combs through the documents, returns a subset, and then attempts to answer the “question” posed by the user using passages from one or more relevant passages from articles in the subset. COVID-19 search also uses abstractive summarization, allowing the search tool itself to generate a short summary of a given document or set of documents.
“COVID-19 Search is designed to serve those on the front lines of medicine and policymaking to accelerate the search for effective vaccines and treatments,” Esteva and Kale wrote. “CORD-19 and TREC-COVID are just the beginning. The computer science community is highly collaborative and we will continue working together and sharing our research to help the larger community develop better search engines for this pandemic and for future challenges.”
To try out COVID-19 Search for yourself, follow this link.