Watson & Natural Language Processing
A John Henry story for the information age, except the humans lose. Ben Zimmer has a summary of the competition between IBM’s Watson computer and two Jeopardy! champions, Ken Jennings and Brad Rutter.
Programming a computer to interpret and respond appropriately to normal human language is really a significant development. Granted that the format of Jeopardy! questions is highly formulaic (among other things, they are pre-classified into categories and each question contains at least two clues) and it’s far from being able to engage in conversation, but it’s still quite impressive.
I’m not going to recount the duel—Zimmer does that quite well—but I will point out one of Watson’s flubs that has personal relevance to me as an American living in Toronto. In the category of “US Cities,” the question, or rather answer, was “Its largest airport is named for a WWII hero. Its second largest for a WWII battle.” Watson buzzed in with, “What is Toronto.” (The correct answer is, of course, Chicago.) When asked how the computer might make such an elementary error, to wit, thinking Toronto was a US city, one of Watson’s programmers explained that since Toronto’s baseball team, the Blue Jays, plays in the American League, the computer might have inferred the city was in the US. Evidently thinking like a human is more subtle than it may seem.
Copyright 1997-2015, by David Wilton