If you are traveling to East Asia, you may soon be able to turn to your mobile phone and use it to obtain on-the-spot quick, easy and cheap translation if you use a new application from Toshiba. This development follows the announcement by Fuji that they will manufacture glasses with translation capabilities (reported in Pangeanic’s blog in November from switched.com).
Toshiba has now developed a trilingual translation system with voice recognition and synthesis. It is compact and light enough to be installed in a cell phone. Unlike the glasses or other existing applications for the iPhone, the software doesn’t offload processing to a more powerful server on the network but performs the task inside the phone. This speeds the whole process and avoids potentially costly data roaming charges if you are using the service abroad.
The software is a reduced version of a PC application that Toshiba already markets in Japan. The program can translate freely between Japanese, English and Chinese and the idea is now that Toshiba can now offer a translation system for cell phones. It first employs speech recognition to determine the language and what has been said. In a case described as “hybrid” by machine translation vendors, the desktop application PC software uses both statistical and rule-based machine translation. By deciding to use either technique (details have not been disclosed on how), the program parses the sentence and break it down into its components. The translation is then performed and a speech synthesis engine outputs the result. This concept would be a market breakthrough into those 3 languages and also a usable concept for other developments elsewhere.
According to reports from computer magazines like computerworld, and pcworld, the version uses statistical machine translation to work on on probability and word order to determine the construction of a sentence while the rule-based method makes use of a database of hundreds of sentences and phrases to extract the meaning. Our team has not been able to find the software as bundled or offered with any Toshiba PC or laptop, though (at least outside Japan).
Moving these capabilities to a cell phone must have been a challenge, current technology in mobile devices lacks the processing power to run efficient translation system. Therefore, reports from news sources say that statistical machine translation is used for English -> Chinese and Chinese -> English and Japanese while rule-based machine translation is used for other combinations.
Martyn Williams from IDG News Service reported that in tests the software worked well. As it happens with any speech recognition software, it is prone to misinterpreting certain words. During tests, this gave rise to incorrect translation. The only solution was obviously to try again. Nevertheless it got things right on some occasions and the uttered sentenced was recognized and also successfully translated.
Toshiba faced several initial challenges when developing the software. Tests were run in quite a powerful smartphone like the TG01, which features a 1GHz Qualcomm processor. The software did manage to execute well although this phone runs Windows Mobile and that imposes a 32MB limit on the size of each process. A development with other language pairs may mean an interesting move by another player outside the industry into translation and offering it as an ubiquitous add-on service, a vision shared by many in the industry.
Commercialization is expected by mid 2010 as work on the application at Toshiba’s research department is almost complete . Interestingly, the Windows Mobile application could be modified to run on other OS, such as Google’s Android or Apple’s iPhone, although no plans have been decided.