NUS Researchers Develop Useful Speech Translation Application
31 January 2011
"Using automatic speech recognition, the system converts the Mandarin speech into Chinese words. The Chinese text is in turn converted to English text. Lastly, the English text is voiced out."
Assoc Prof Ng Hwee Tou, NUS School of Computing
HOW IT WORKS: The iPhone application devised by the NUS team can perform language translation from English to Mandarin and vice versa as well as from Malay to English
THE BRAINS BEHIND THE NEW APPLICATION: (Front) Assoc Prof Ng Hwee Tou (left) and Dr Sim Khe Chai. (Back from left) PhD students Mr Wang Pidong, Mr Wang Xuancong, Mr Hoang Trong Nghia and Research Fellow Dr Preslav Nakov. Missing from the photo are PhD student Mr Liu Chang and Research Fellow Dr Lu Wei
Travellers will find the iPhone application on speech translation by researchers from the NUS School of Computing useful when they go overseas. Developed by Assoc Prof Ng Hwee Tou and Dr Sim Khe Chai, the novel iPhone application can translate speech from English to Mandarin and vice versa as well as from Malay to English. The iPhone application is undergoing testing before it is released.
Assoc Prof Ng explained how a user who speaks a sentence in Mandarin can hear the translated English version. “Using automatic speech recognition, the system converts the Mandarin speech into Chinese words. The Chinese text is in turn converted to English text. Lastly, the English text is voiced out,” he said. All this is done within a matter of seconds.
On what led them to work together on this application, Assoc Prof Ng shared that they naturally delved into the area of speech translation given his expertise in natural language processing combined with Dr Sim’s proficiency in speech processing.
Their research is aided by a funding of S$4 million over a period of five years by Singapore’s Media Development Authority. Started in mid-2008, the work is carried out in collaboration with CSIDM (China-Singapore Institute of Digital Media), the first R&D institute of the Chinese Academy of Sciences Institute of Automation set up outside China.
According to the researchers, there are currently three similar iPhone applications that can perform speech translations but none of them translates Malay speech. Once released, the NUS application might be offered free of charge. The revenue model for the application has yet to be finalised.
Already, Assoc Prof Ng and Dr Sim have started working on improving their iPhone application. While the current application supports English, Mandarin and Malay, more languages especially from the region can be incorporated.
In December 2009, the NUS team won second place in a Chinese-English spoken language translation task, an international evaluation which was conducted as part of the International Workshop on Spoken Language Translation.