The automatic translation software boasts a vocabulary of 50,000 Japanese and 36,000 Chinese words and realizes automatic speech-to-speech interpretation of travel conversation. A wide range of travel conversation used in a variety of travel situations was taken into consideration during the development of the software, and it has principally been enabled by the realization of speech recognition of general conversation of standard spoken Chinese. Although standard spoken Chinese is the same nationwide, accents differ according to region, making speech recognition complex. In order to combat this issue, data of spoken conversation was gathered from a large number of people from different regions in China, allowing the creation of a large-scale Chinese-conversation speech corpus.

The software includes a compact translation engine for translation of spoken words, an advanced wave-concatenative speech synthesis engine, which realizes high-performance reading through an advanced, wave-concatenative speech synthesis method based on a wide range of speech data and a total integration module that controls collaborative operation of the speech recognition engine, the machine translation engine and the speech synthesis engine, realizing real-time, automatic Japaneseâ†‿Chinese speech translation on a PDA.

In 1999 NEC released its automatic Japaneseâ†‿English translation software for notebook PCs and in 2002 the PDA-operational Japaneseâ†‿English travel conversation, automatic speech translation software. The speech-to-speech translation software capable of being processed on low-power-consumption multi-core processors, optimal for small devices such as mobile phones was released in 2005. Although NEC’s automatic speech translation software research has been focused on [1] Japaneseâ†‿English translation to date, NEC has been able to realize Japaneseâ†‿Chinese speech translation software by installing a newly created large-scale Chinese language conversation speech corpus, a Chinese travel conversation text corpus, and an acoustic model and a language model that were developed based on both of these, in the compact, high speed speech recognition engine that NEC has been able to create owing to many years of research in this area.