Advanced Speech Technology Laboratory

The Advanced Speech Technology Laboratory will realize practical speech recognition technologies in 10 languages (Japanese, English, Chinese, Korean, Thai, Vietnamese, Indonesian, Myanmar, Spanish, and French) with the aim of implementing them in society for the 2020 Tokyo Olympics and Paralympics. In doing so, the following research and development activities will be conducted: (1) building speech corpora of about 2,000 hours for each of the 4 languages, Japanese, English, Chinese and Korean, and about 500 hours each for the other languages; (2) developing multilingual and multidisciplinary language models; and (3) developing high-speed, high-accuracy speech recognition engines. The Laboratory will also conduct research and development of speech synthesis technologies to realize practical speech synthesis systems in the above 10 languages. Looking ahead to the world post-2020, the Laboratory will conduct research and development of (4) technologies to recognize speech generated by multiple people speaking different languages in environments such as public spaces with background noise and echoes, with the aim to realize technologies that convert all speech contents around the globe into text; and lastly (5) technologies for mixed-language dialogues in multiple languages.