Sign In

Communications of the ACM

ACM TechNews

Google's 'babel Fish' Heralds Future of Translation

View as: Print Mobile App Share:
Ashish Venugopal

Google translation researcher Ashish Venugopal

Credit: TechCentral

The Google Translate project seeks the enablement of real-time language translation by building "statistical models that are automatically training themselves and learning all the time," says Google Translate researcher Ashish Venugopal. "As people translate new content on the Web, our systems pick this up and it adds the words."

Venugopal notes that Google Translate currently supports 63 languages. He says the idea is not to tell the computer how to translate each sentence, but rather feed it general patterns to search for.

"When it sees new data, it uses those patterns, matches that to data, and then comes up with a model that it uses to translate sentences," Venugopal says. He concedes that the statistical strategy can lead to scenarios in which the inputted data generates a bizarre translation, and the Google Translate researchers are attempting to limit such situations.

Venugopal predicts that machine translation will reach a point where it can effectively address 80 percent of the use cases "in a reasonably short time." However, he cautions that "the last 20 percent will be incredibly hard."

Venugopal also notes that real-time voice translation using devices such as mobile phones is already possible, but not on a simultaneous basis.

From TechCentral (South Africa)
View Full Article

Abstracts Copyright © 2012 Information Inc. External Link, Bethesda, Maryland, USA 


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account