Power > Business Strategy > BI and Analytics

Watson Goes to Work

Not content to rest on its ‘Jeopardy!’-winning laurels, Watson moves into the healthcare arena

You’d think that after its big win on “Jeopardy!” last year Watson would be lounging on a beach somewhere drinking Mai Tais.

Well, the brainiac Watson is hardly one to let its international fame go to its head. No, it decided instead to broaden its education and now, thanks to the equally great minds behind it, use its natural-language capabilities to assist people in a wide variety of fields.

One of the first of these is healthcare. Beginning early this year, IBM and WellPoint Inc., one of the largest health-insurance providers in the U.S., will be piloting a program using the DeepQA technology that allows Watson to answer medical questions using natural language.

So instead of answering questions such as “Who was Richard Nixon?,” Watson will be delving into topics like “What’s the best treatment for this type of pancreatic cancer?” This will let both the insurance company and its member physicians chart the most effective course of action for a long list of maladies. Physicians can also enter personal patient information and symptoms, and Watson will scour medical journals and other sources for near-instantaneous results.

“It seemed like medical diagnosis and treatment options were a natural extension of Watson’s capabilities,” says Bill Rapp, chief architect, Watson for Healthcare. “That’s because of the massive amounts of information involved in the industry, and Watson’s very good at ingesting, interpreting and making sense of that information very quickly and returning very accurate results.”

Only the Beginning

Initially, the development of Watson appeared to be a mere a lark—a what-if mental exercise involving natural-language processing on a binary, if-then system. But Watson’s much-heralded “Jeopardy!” appearance turned out to be more a proof-of-concept starting point than an end unto itself. That it piqued the world’s imagination was a welcome bonus.

Even if Watson had lost the game, which was possible, it still would have been a success. After all, it was the first practical example of a system that could understand and respond not only to factual questions but also to more nuanced ones, including puns.

This capability was made possible by its underlying DeepQA architecture. Unlike other QA systems, DeepQA can respond to both structured and unstructured knowledge, although it excels at the latter by creating connections between information that might not otherwise be relatable by traditional QA systems. Hence Watson’s ability to respond to puns. As it receives more of this unstructured information and applies it to a hierarchy of possible answers, it becomes increasingly confident in its final answer.

Of course, that’s only part of the story. Watson is also capable of processing enormous amounts of data by running thousands of analytics algorithms simultaneously in a massively parallel processing environment. The speed of its underlying POWER7* technology was in part why Watson could ring in so quickly during the “Jeopardy!” tournament.

So while Watson, with its DeepQA “brain” and capability to process huge amounts of data in seemingly no time, was developed initially to become a viable “Jeopardy!” contender—if not an outright winner—it became increasingly clear to IBM that the technology could be used elsewhere, wherever natural-language applications might come in handy. Healthcare was a logical choice.

“There’s a very large amount of medical-related information already out there, and it’s growing all the time. In fact, some sources say it doubles in size every few years, and as a result, it’s becoming increasingly difficult for physicians and other care providers to keep up with it,” Rapp says. “This is where Watson-based technology can come into play.”

“We don't have to be prepared to answer any question in any category, as was the case with the ‘Jeopardy!’ Watson. Now we can focus specifically on questions such as ‘What disease is this?’ ” —Eric Will, lead architect, Watson for WellPoint Solutions

Tweet