IBM breaks the language barrier

US Joint Forces Command has partnered with IBM to supply US forces serving in Iraq with conversational English to Arabic "speech-to-speech" translators, to help combat the shortage of Arabic linguists.

US Joint Forces Command

(USJFCOM) has partnered with

IBM

to supply

US

forces serving in

Iraq

with conversational English to Arabic "speech-to-speech" translators, to help combat the shortage of Arabic linguists.

The Multilingual Automatic Speech-to-Speech Translator, or MASTOR, software will initially be deployed on 35 ruggedised laptops to Army Medical, Army Research Lab, Army Intelligence, Special Operations Command and the Marine Corps.

They will be used to facilitate military and medical-oriented conversations with members of the Iraqi security forces, in hospital settings and during daily interactions with Iraqi citizens. Some of the devices will also be used in the US to help train military personnel in advance of deployments.

Currently, commercially available translation systems can only work with pre-programmed fixed phrases. MASTOR offers the ability to hold a free-form conversation without having to memorise any pre-determined phrases.

The goal of MASTOR is to convey the meaning of what is said, even if the speaker or speech recogniser makes minor errors. During operation, the user speaks into a microphone, MASTOR recognises and translates the speech, then vocalises the translation in the target language for the foreign language speaker to hear. The foreign language speaker can then speak into the microphone in their own language, and MASTOR translates and vocalises their speech back to the original language. MASTOR produces an audible and text translation of the spoken words that can run on a PDA, tablet PC or laptop computer.

During development IBM combined automatic speech recognition, language understanding and synthesis technologies. The tight coupling of speech recognition and understanding effectively mitigates the impact of speech recognition errors and non-grammatical inputs common in conversational colloquial speech on the quality of the translated output, resulting in a highly robust system.

MASTOR is available in two-way English to Iraqi Arabic, English to Modern Standard Arabic and English to Mandarin Chinese; additional languages are planned.