Today's Hard|Forum Post
Today's Hard|Forum Post

Wednesday April 05, 2017

Google Brain Empowers Neural Networks to Deliver a more Accurate Translation Service

The Google Brain team, a deep learning research arm of the company, has levied the power of neural networks to breakdown the barriers to a more accurate translation service. Other services such as Skype have to transcribe the incoming spoken language to written text and then read the transcription aloud in the desired language. This leads to a lot of unwanted errors as there are many points at which errors can be introduced into the translation with this methodology. Google Brain is different because it is an A.I. trainer that listened to hundreds of hours of Spanish audio and correlating it to the equivalent chunks of English text. Then when it hears spoken Spanish; it alters the waveform until it matches the audio it has previously associated with chunks of translated English text. According to Annette Choi at PBS, "It’s the computer equivalent of your ears hearing Spanish while your brain understands the words as English." By cutting out the Spanish audio to Spanish text "middleman" it is much more accurate than conventional translators using the BLEU quality standard for machine language translators.

News Image

The practical use of this technology could create translators for languages in emergencies such as floods, hurricanes, etc. An example of this need would be for the Haitian Creole language after a recent hurricane as the aid workers couldn't communicate with the local populace to assist them. In some countries there are only 100 members of a tribe that speak an unknown language. This technology could create a faster translation platform to communicate with them. Reading the original paper on the work yielded this possible future use of the technology, "An interesting extension would be to construct a multilingual speech translation system following [32] in which a single decoder is shared across multiple languages, passing a discrete input token into the network to select the desired output language."

International disaster relief teams, for instance, could use it to quickly put together a translation system to communicate with people they are trying to assist. When an earthquake hit Haiti in 2010, says Goldwater, there was no translation software available for Haitian Creole. Goldwater’s team is using a similar method to translate speech from Arapaho, a language spoken by only 1000 or so people in the Native American tribe of the same name, and Ainu, a language spoken by a handful of people in Japan.

Discussion