New Website Transfix.it Uses Crowdsourcing to Improve Computer Language Translation
San Diego, CA, August 10, 2012 --(PR.com)-- A new website is aiming to improve online language translation by letting its users correct computer translations when the software comes up short. Transfix.it, the brainchild of founder and CEO Michael Messner, is “not just an evolution, but a revolution in how we translate language online,” according to Messner.
Since the 1980s, when language translation software was in its infancy, translation software has relied on a combination of translation dictionaries and already-translated transcripts as source material. Google Translate, for instance, uses transcripts from United Nations speeches that have already been accurately translated into multiple languages. “What software never really gets, though, is context,” says Messner. “Take the word ‘dust,’ for instance. If I say I’m dusting a bookshelf, you’d understand that I mean I’m removing dust. If I say I’m dusting a cake with powdered sugar, you know I mean the opposite. But a computer doesn’t.” This lack of context is often the cause of the amusing translation errors that software like Google Translate and Babelfish can make.
The way that Transfix.it solves this problem is by first translating text by using software, and then letting the users of the site fix the errors that the software makes. This method of translation, known as “computer aided translation,” or CAT, has been used by professional translators for many years. “The software can get you about 75% of the way there, but it takes a human to get you to 100% accuracy,” says Messner.
The translators on the site work without compensation, much like the editors on Wikipedia. Many social networks count on their users to edit each other’s work and regulate each other’s behavior. One high-profile translation website that was recently launched, Duolingo, leverages its user base to translate entire websites, one sentence at a time. The translators on Duolingo are people taking language lessons. The accuracy of their translations is then voted on by other students.
Transfix.it caters to a different audience, according to Messner. “Duolingo helps web developers translate entire websites. We’re focusing on short passages of text like emails, song lyrics, forum posts, and text messages. Also, the translators on Duolingo are language students trying to learn the language that they’re translating, and are translating from scratch. They’re not native speakers of the language, almost by definition. In our early tests we’ve found that native speakers are more than willing to help translate simply for the benefit of the community and some recognition on the home page.”
One danger to all of these sites is the advancement of computer translation algorithms and processing power. Is there concern that Transfix.it will be obsolete by the time it attracts a crowd of users? “I truly believe that language translation is similar to weather prediction,” says Messner. “In the 1960s meteorologists thought that perfectly predicting the weather would soon just be a matter of gathering enough data and running it through computers with enough processing power. Huge advances in data collection and processing led to only minor improvements in their predictions, though. I think that computer translation will continue to improve, but we are still a long, long way from software translation that rivals a human.”
Until then, sites like Duolingo and Transfix.it will bet on the power of the crowd over the power of the computer.
Michael Messner can be contacted at info@transfix.it or by phone at 619-813-9209.
Since the 1980s, when language translation software was in its infancy, translation software has relied on a combination of translation dictionaries and already-translated transcripts as source material. Google Translate, for instance, uses transcripts from United Nations speeches that have already been accurately translated into multiple languages. “What software never really gets, though, is context,” says Messner. “Take the word ‘dust,’ for instance. If I say I’m dusting a bookshelf, you’d understand that I mean I’m removing dust. If I say I’m dusting a cake with powdered sugar, you know I mean the opposite. But a computer doesn’t.” This lack of context is often the cause of the amusing translation errors that software like Google Translate and Babelfish can make.
The way that Transfix.it solves this problem is by first translating text by using software, and then letting the users of the site fix the errors that the software makes. This method of translation, known as “computer aided translation,” or CAT, has been used by professional translators for many years. “The software can get you about 75% of the way there, but it takes a human to get you to 100% accuracy,” says Messner.
The translators on the site work without compensation, much like the editors on Wikipedia. Many social networks count on their users to edit each other’s work and regulate each other’s behavior. One high-profile translation website that was recently launched, Duolingo, leverages its user base to translate entire websites, one sentence at a time. The translators on Duolingo are people taking language lessons. The accuracy of their translations is then voted on by other students.
Transfix.it caters to a different audience, according to Messner. “Duolingo helps web developers translate entire websites. We’re focusing on short passages of text like emails, song lyrics, forum posts, and text messages. Also, the translators on Duolingo are language students trying to learn the language that they’re translating, and are translating from scratch. They’re not native speakers of the language, almost by definition. In our early tests we’ve found that native speakers are more than willing to help translate simply for the benefit of the community and some recognition on the home page.”
One danger to all of these sites is the advancement of computer translation algorithms and processing power. Is there concern that Transfix.it will be obsolete by the time it attracts a crowd of users? “I truly believe that language translation is similar to weather prediction,” says Messner. “In the 1960s meteorologists thought that perfectly predicting the weather would soon just be a matter of gathering enough data and running it through computers with enough processing power. Huge advances in data collection and processing led to only minor improvements in their predictions, though. I think that computer translation will continue to improve, but we are still a long, long way from software translation that rivals a human.”
Until then, sites like Duolingo and Transfix.it will bet on the power of the crowd over the power of the computer.
Michael Messner can be contacted at info@transfix.it or by phone at 619-813-9209.
Contact
Transfix.it
Michael Messner
619-813-9209
transfix.it
Michael Messner
619-813-9209
transfix.it
Contact
Categories