
The great linguistic distance of Japanese from other major commercial languages has stimulated a great deal of MT-related activity in Japan but also continues to present challenges for translation quality. One could sense at the TAUS meeting a strong interest in Japan in the emerging new generation of MT technologies and that perhaps the impetus to experiment with and implement new solutions is starting to overtake the longstanding cautious skepticism towards MT that has prevailed in Japanese companies.
On April 15 and 16 TAUS held its first Executive Forum in Japan, hosted at Oracle’s Training Center in the Aoyama Center. The Forum was attended by 40 people, mostly delegates from Japanese corporations, both buyers and suppliers of translation services and technologies. MT research has been thriving in the sixties and seventies of last century. Professor Hitoshi Isahara – President of the AAMT – shared an overview of the MT landscape in his keynote opening to the forum. Statistical approaches to MT are now slowly being adopted in Japan, among others through the work done by the National Institute of Information and Communication Technology (NICT). Of course the success of Google, Microsoft, Language Weaver (all following the statistical route) does not remain unnoticed in Japan.. In a series of technology briefings the participants familiarized themselves with the latest developments at Fujitsu and Toshiba (two of the Japanese MT veterans), from Tokyo University (affiliated with NICT), from SDL and Language Weaver and finally from Pangeanic, a language service provider that embraced the open-source Moses SMT system to provide its customers with customized MT solutions.
As the post-forum survey among all participants the biggest interest was in the MT use cases, of which there were many at the TAUS Forum. We learned from Cisco’s use of the Systran MT engine to translate support content and its first experiences with statistical post-editing. We heard about Sun’s experience of evaluating several MT systems, followed by the presentation by CA who experienced more than a 60% drop in translation cost using the Toshiba Honyaku technology. Pangeanic raised a lot of interest talking about its pilot project for Sony Europe, using the Moses statistical MT engine and translation data from the TAUS Data Association repository. In just two years, Adobe has moved in a big way into using MT, showing productivity gains between18.5% and 41.5%, covering four languages and using a traditional and a statistical engine. Japanese is a new MT language for Adobe, for which Language Weaver is the partner. City University Hong Kong took us into some of statistical routines underlying its use of MT technology for the translation of patent applications, an area of incredible growth in China. Finally, hot from the press, participants were presented an old idea with a new twist: the use of interlingua techniques to resolve some of the particular challenges in English-Japanese translation. The inventor – ITAS corporation – promised to keep us informed about future advances.
The TAUS Forum also zoomed in on best practices in post-editing. The TAUS overview was enriched with valuable lessons learned from CA and Language Weaver among others. On Friday afternoon participants discussed innovation and change scenarios, inspired by Lionbridge’s launch of the Translation Workspace and adoption of cloud-based translation memories. The alliance with IBM was still a secret at the time of the TAUS Tokyo Forum. Jonckers asked the audience to think about localization business structures undergoing fundamental changes in recent years: localization companies do not need to have offices anymore in every major city to service global customers, or do they….?
In summary, all participants agreed that the TAUS Forum served a tremendous need for the translation community in Japan to meet and exchange experiences. A dedicated platform for this was missing. Therefore TAUS made the commitment to be back next year: block the dates April 14 and 15, 2011.
Please find all presentations in the member section.
![]()


