Super-function based Japanese-English machine translation

摘要

We have presented a new method called super-function based machine translation (SFBMT). SFBMT uses super-function (SF) to translate without syntactic and semantic analysis as most conventional MT systems do. SF represents correspondence with a source language sentence structure and a target language sentence structure directory. According to this feature, the system realizes very fast translation. Moreover we do not feel the sense of incompatibility in the translation result because the SFs are automatically generated from a large corpus. We describe (1) SF based machine translation, (2) extracting SF, (3) the evaluation experiment. As the experimental results, we obtain 84% of effective rates of translation.