The American group Google develop a program for first mobile phone able to instantly translate a foreign language during the call, and the system could be operational in the next few years, according to newspaper The Times.

Developing existing technologies of voice recognition and translation, Google could bring major changes communication between the over 6,000 languages spoken worldwide.

Google has created already an automatic translation of text for computer, which is constantly improved by scanning millions of websites and documents in different languages. PC program covers, at present, 52 languages, the latter Haitian Creole, added last week.

Also, the group also owns a voice recognition system that allows mobile users to search the Internet without using keys.

IT giant is working to combine the two technologies to obtain a software able to recognize the voice of a person and transform it into a synthetic play in a foreign language. As a human interpreter, the program will analyze longer sequences of speech, listening to understand the meaning of sentences and phrases before attempted a translation.

“We believe that voice translation will be possible and it will be acceptable in a few years. To go well, it takes a combination of translation and voice recunasterea, both with high accuracy, exactly what we are now developing“, said head of Translation by Google, Franz Och.

If the automatic translation of text is acceptable efficiency, development of voice recognition software presents several difficulties.

“Everyone has the voice, accent and tone different. But recognition should work on mobile phones, because they are personal. Cell phone owner should know the voice of voice search previously completed,” added Och.

Translation software will become more accurate the more will be used more. Also, while some of sitemelor translation using simple rules of grammar, Google uses its vast base of websites and translated documents to improve the accuracy of the system.

“The more gather more data, so get a higher quality,” he said.

Some experts believe that translation difficulties remain high during the call.

“The problem with voice recognition is the diversity of accents. No system can not cope with something now. Maybe Google will get results faster, but I think that is unlikely to have in the years to a device can handle dialect, spoken quickly, the Glasgow region, “said David Crystal, professor at Bangor University in Wales.