(PhysOrg.com) -- Google is developing a translator for its Android smartphones that aims to almost instantly translate from one spoken language to another during phone calls.
Head of translation services at Google, Franz Och, said he believed almost instant speech-to-speech translation should be possible if the accuracy of voice recognition and machine translation can be improved. He said Google is working on this, and he expected the technology to “work reasonably well” in a few years. There has been a great deal of progress in voice recognition and machine translation in recent years, especially the latter, thanks to funding from the U.S. Defense Advanced Research Projects Agency (DARPA).
Google's system would analyze speech in the same way as a human interpreter, listening to a package of speech to gather the full meaning before it attempts to translate. It will improve in accuracy the more it is used.
Google already has a website translation program for 52 languages, with an expanding database that continually improves its accuracy. It also has a voice recognition application for smart phones to allow users to search the web by speaking their commands into the phone. The new system will combine the two technologies, and will use the database Google has built up by crawling websites in different languages to improve its understanding of them.
Translation of speech will provide much bigger challenges than text translation, however. Mr Och said the task is difficult because everyone has a different voice, pitch, and speed of speaking. The huge variety of accents, dialects and colloquialisms also presents problems that some linguistics experts, such as David Crystal, honorary professor of linguistics at Bangor University in Wales, consider insurmountable. If it does prove possible, and becomes available to the average user, it may help preserve languages and avoid a globalized language monoculture.
Google expects to release a “basic version” of the translation application in a couple of years. Apple’s iPhone already has a speech translator called Jibbigo, which is available for English/Spanish and English/Japanese.
Explore further: Software provides a clear overview in long documents