Toshiba has developed a new hybrid translation method
that realizes a speech-to-speech translator for all six
translation directions among the Japanese, Chinese, and
English languages. This hybrid method enables the system
to produce natural and accurate translations for arbitrary
spoken and recognized utterances. Two of Toshiba's
original complementary machine translation (MT)
methods are utilized: forest-driven rule-based MT, and
chunk- and stochastic-type example-based MT.
Several field tests of this system were performed in
Japan, China, and Australia in 2008, with participation
by more than 130 native-speaking subjects for these
three languages in total. The results of these field tests
demonstrated that about 70% of typical tasks in overseas
travel situations can be achieved within about two
minutes.
To Top