Development and application of multilingual speech translation

This paper describes the latest version of handheld speech-to-speech translation system developed by National Institute of Information and Communications Technology, NICT. As the entire speech-to-speech translation functions are implemented into one terminal, it realizes real-time and location free speech-to-speech translation service for many language pairs. A new noise-suppression technique notably improves speech recognition performance. Corpus-based approaches of recognition, translation, and synthesis enabled wide range coverage of topic varieties and portability to other languages. Currently, we mainly focus on translation between Japanese, English and Chinese.