This week the New York Times published a fascinating in-depth piece about Watson, a new supercomputer created by IBM specifically to answer the type of vague, allusive riddles posed on the quiz show “Jeopardy!” The show may even host the supercomputer as a contestant as early as fall 2010. Anyone interested in search technology, and specifically natural language query processing, should definitely give this article a look.
It’s not just that Watson can answer simple, factual questions – it’s that the machine can even put together two concepts, such as a typical Jeopardy! question that requires contestants to name a combination between a candy bar and a Supreme Court Justice – “Baby Ruth Ginsberg.” There are no major breakthroughs in search technology to report; instead, the machine is able to answer the questions in the short time frame given to contestants by virtue of its enormous computing power and memory. But it doesn’t just use one single search algorithm to answer questions. Instead, Watson analyzes its knowledgebase using many different statistical methods, generating various possible answers. If multiple methods pinpoint the same result, it has a higher chance of being correct, and will get a higher “confidence” score from the computer. After a certain item passes an acceptable confidence threshold, Watson will buzz in and answer the question (of course, the machine will use the famous “What is ___ ?” format that the game show requires).
After checking out the article, you can also play an interactive trivia game, simulating a round of “Jeopardy!” with Watson. I was amazed at the questions Watson was able to answer, but just as interested to learn about the questions it missed, and why.
The obvious practical application for Watson’s technology would be providing quick decision support. For example, a later generation system might analyze the body of medical research to supply fast answers for doctors, or to provide the first line of customer support at a virtual call center. For patent and prior art searching, a program that can produce a quick match is always interesting, but when it fails, we’re often still in the position of having to prove a negative. However, we also stand to benefit from any improvements in search technology that arise from Watson’s development.
What are your ideas about how Watson’s technology might one day benefit the IP industry? Let us know in the comments!
This post was contributed by Intellogist team member Kristin Whitman.