Bangkok--30 Jul--NECTEC
The First Asian Network-Based Speech-to-Speech Translation System The Asian Speech Translation Advanced Research Consortium (A-STAR), aiming at the realization of the network-based speech-to-speech translation systems in Asian region, was established in June 2006 by NICT/ATR (Japan). A-STAR originally started with 6 members: Japan (NICT/ATR), Korea (ETRI), Thailand (NECTEC), Indonesia (BPPT), China (NLPR-CASIA), and India (CDAC). In 2008, Vietnam (IOIT) and Singapore (I2R) also joined A-STAR.
A-STAR was founded in order to create a basic infrastructure for spoken language communication for overcoming the language barriers in the Asia-Pacific region. The consortium is working collaboratively to collect Asian language corpora, to create common speech recognition and translation dictionaries, to develop Web service speech translation modules for Asian languages, and to standardize interfaces and data formats for connecting speech translation modules internationally. The consortium has moved to create an expert group within APT ASTAP (the Asia- Pacific Telecommunity Standardization Program) in order to develop a draft for a standardized interface and data format for connecting speech translation modules to enable their connection over the Asian-Pacific region through the Internet. The research activities have also been adopted as the APEC TEL (Telecommunications and Information) project.
On July 29, 2009, A-STAR launched “the first Asian network-based speech-to-speech translation system” that can perform real-time, location-free, multi-party communication between speakers of different Asian languages. The key points are:
- Eight A-STAR member research groups joined the experiments covering 9 languages which
included eight major Asian languages (Japanese, Chinese, Korean, Thai, Indonesian, Malay,
Vietnamese, Hindi) and additionally the English language.
- Each A-STAR member provided a portion of the spoken language technologies: automatic
speech recognition (ASR), machine translation (MT), and text-to-speech (TTS) through STML
web servers. Currently, the system can perform ASR for 8 different languages, TTS for 9
different languages, and MT for 72 different language pairs.
- The client applications are implemented on a handheld mobile terminal device (VAIO), which
allows portable speech-to-speech translation.
- Any client user can access, in real time from anywhere, all provided A-STAR ASR/MT/TTS
STML servers, and can perform language translation between up to 4 parties at the same
time. There are 9 possible languages giving rise to a possible 246* different multi-party
speech-to-speech translation language combinations.
- The system domain covers 160,000 travel expressions, plus additional named entities (NE)
from major Asian countries (e.g., tourist areas: Bulkuksa-Korea, Watprakaew-Thailand;
attractions: Wayangkulit-Indonesia, Khatak-India, etc).
The consortium is continuing their activities internationally and partners are still being sought for languages not only in Asia, but around the world. This is expected to accelerate the speed of research towards the realization of practical speech translation systems.