Watson Takes on Science

IBM’s supersystem gives the research scientist encyclopedic knowledge of all papers written on a specific topic.

IBM has announced advances in their Watson Discovery Advisor cloud service, which they launched earlier this year, to bring the power of Watson to assisting research scientists. Watson is IBM’s supercomputer system that became famous for winning Jeopardy. In January, IBM unveiled a US$1 billion investment in developing and extending Watson to find practical applications for its impressive capabilities. The Discovery Advisor is part of that effort.
According to IBM, “Building on Watson’s ability to understand nuances in natural language, Watson Discovery Advisor can understand the language of science, such as how chemical compounds interact, making it a uniquely powerful tool for researchers in life sciences and other industries.”
Scientists in academic or commercial research centers can deploy Watson to analyze and test hypotheses rapidly using data in millions of scientific papers. To win at Jeopardy, Watson read and organized data from thousands of books and other written documents, and programmers successfully taught it to extract useful information to address specific questions. Applying the same principle to science, the Watson team is focused on turning Watson loose on the thousands of academic papers written on a specific subject. More than a million scientific papers are published every year. IBM quotes the National Institutes of Health with the observation that “a typical researcher reads about 23 scientific papers per month, which translates to nearly 300 per year, making it humanly impossible to keep up with the ever-growing body of scientific material available.”
Watson, on the other hand, can consume all the available information and look for connections and correlations that might not occur to a human. The press release quotes the example of a study by the Baylor College of Medicine and IBM, in which scientists used Watson technology embedded in the Baylor Knowledge Integration Toolkit to analyze 70,000 scientific papers on a specific protein related to many cancers and identify six proteins that appear promising for further research.