The system does real-time network-based speech translation. |
The Institute for Infocomm Research (I²R), a member of the Singapore Agency for Science, Technology and Research (A*STAR) family, has invented a technology that can provide speech-to-speech translation between Malay and eight other languages used in Asia. These include Japanese, Chinese, Korean, Thai, Indonesian, Vietnamese, Hindi and English.
The technology was developed over two years by I²R’s Speech Translation team led by Dr Li Haizhou, who heads the Human Language Technology department. Members include Dr Ma Bin, Ms Aw Ai Ti, Dr Vladimir Pervouchine, Ms Tong Rong, Mr Bai Shuanhu and Ms Sharifah Mahani Aljunied.
The system operates with the help of a portable device and does real-time network-based speech translation. The 72 possible speech-to-speech translation combinations were demonstrated by the Asian Speech Translation Advanced Research Consortium at the launch of the translation system in July.
Each of the research groups participating in the demonstration contributes a portion of the spoken language technologies for automatic speech recognition, machine translation and text-to-speech translation. Currently, the system can perform automatic speech recognition and text-to-speech synthesis for nine different languages, and machine translation for 72 different language combinations.
The client applications are implemented on a handheld mobile terminal device, which allows portable speech-to-speech translation. Any client user can access the system in real time from anywhere in the world and perform multi-party speech translation, provided the mobile device runs a speech-to-speech translation client and the system supports the Speech Translation Markup Language.
The system domain covers 160,000 travel expressions, as well as named entities from major Asian countries. Examples of these named entities include tourist areas such as Bulkuksa-Korea and Watprakaew-Thailand, and attractions such as Wayangkulit-Indonesia and Khatak-India.
Professor Lye Kin Mun, Deputy Executive Director (Research) said, “I²R has come a long way in developing speech recognition and machine translation technologies. Today’s demonstration demonstrates our technological capabilities working with other Asian partners in breaking down communication barriers. We hope to work with more countries on a global scale with this imminent success.”
I²R, representing Singapore, joined the Asian Speech Translation Advanced Research Consortium last year, together with the Vietnam Institute of Information Technology.
The consortium was established in June 2006 to create a basic infrastructure for spoken language communication to overcome the language barriers in the Asia-Pacific region.
Founding countries included Japan (National Institute of Information and Communications Technology and the Advanced Telecommunications Research Laboratories), Korea (Electronics and Telecommunications Research Institute), Thailand (National Electronics and Computer Technology Centre), Indonesia (Badan Pengkajian Dan Penerapan Teknologi or The Agency for the Assessment and Application Technology), China (National Laboratory of Pattern Recognition of the Institute of Automation of Chinese Academy of Sciences), and India (Centre for Development of Advance Computing).
The consortium is collaborating to collect spoken records of Asian languages, create common speech recognition and translation dictionaries, develop Web service speech translation modules for Asian languages, and standardise interfaces and data formats for connecting speech translation modules internationally.