Broad-spectrum natural language processing

Microsoft's Rick Rashid is talking about broad-spectrum natural language processing. In terms of voice recognition, Asian languages have crossed the threshold where voice beats typing as a mode of input. Not yet true for English yet, he admits.

But there's more natural-languages smarts in Office than we might suspect, Rashid says. Inside the office grammar checker, there's deep parsing of text. This information is extractable, Rashid says. You can ask the DLL to give you a parse tree, and can look at deep structure: subject, object. I'd like to see that!

Next there's the issue of data-mining large bodies of text. Rashid demos the AskMSR system. It knows when Abe Lincoln was born (adjusting for wrong answers on some web pages). It knows that the answer to everything is 42.


Former URL: http://weblog.infoworld.com/udell/2002/05/14.html#a237