New Scientist (subscription required for full text) has a really interesting interview with Google’s Director of Research, Peter Norvig
“His big idea is that if you can amass sufficient data, a few relatively simple statistical algorithms are enough to solve some of the most vexing problems in machine learning, such as automated language translation.”
“Google's translation tools are the most mature product of this approach. "In the past, people had thought of this as being a linguistics problem," says Norvig, which meant designing software that could replicate a human translator's understanding of the languages involved, including their grammatical rules.
Instead, Norvig's team has compiled texts that have already been translated, and applies statistical techniques to train the system to learn translations of unfamiliar words and how they are used in context. "Essentially we are just building this big model of probabilities," Norvig says.”
“Speech recognition is a tougher nut to crack. Because of the almost infinite variety in accent, timbre and other characteristics of human speech, the conventional approach has been to train systems to learn individual users' vocal tics. That's fine if you want a personal dictation machine, but useless for allowing computers to interpret anything that's said to them, by anyone.
Norvig is convinced that speech recognition will fall to the "big data, simple algorithms" approach. The problem is finding enough data, as the spoken word is not represented online as comprehensively as text and images. As we discuss this issue, Norvig makes a revealing admission about the launch of Google Voice, which among other things transcribes phone messages and sends them to your email inbox: "One of the reasons we had this phone service is that we wanted to capture lots of interactions; hear different accents and different voices saying different things."”
I wonder if this approach is actually some kind of commodificaton of knowledge work, but just on an epic scale that we have yet to truly appreciate?
Posted by: twitter.com/PaulSweeney | May 01, 2011 at 01:11 PM
Paul, sure that will be one impact and linguists and medical transcribers among others will be affected. What I like is the big application, machine learning vision of Google - lot's of "big data" vendors have hammers looking for nails.
Posted by: tmirchan | May 01, 2011 at 03:09 PM