Best Paper Award
Development of gVoiceTrah Multi-Lingual Speech Transration System for Practical Use
Shigeki Matsuda, Teruaki Hayashi, Yutaka Ashikari, Yoshinori Shiga, Hideki Kashioka,
Keiji Yasuda, Hideo Okuma, Masao Uchiyama, Eiichiro Sumita, Hisashi Kawai, Satoshi Nakamura
[Trans. Inf. & Syst. (JPN Edition), Vol. J96-D No.10, Oct. 2013]

Shigeki Matsuda

Teruaki Hayashi

Yutaka Ashikari

Yoshinori Shiga

Hideki Kashioka

Keiji Yasuda

Hideo Okuma

Masao Uchiyama

Eiichiro Sumita

Hisashi Kawai

Satoshi Nakamura
       
  In this paper, the system outline, details of each module, and huge-scale practical evaluation of the multi-lingual speech translation system ``VoiceTra'', developed by National Institute of Information and Communications Technology (NICT), are described. VoiceTra consists of a client application and a server system. The client application carries out speech input, indication of translated texts, and play back of synthesized speech. On the other hand, the server system carries out speech recognition, machine translation, and translated speech synthesis. The client application and server system communicate by means of an original communication protocol. The greatest feature of VoiceTra is multilingual translation. Although speech input of VoiceTra is restricted to six languages, text input supports mutual translation among 21 languages. Therefore, people in various countries can enjoy the speech translation services with a light heart. In addition, the user interface of VoiceTra is elaborately designed for easy use of the speech translation system.
  The huge-scale practical evaluation of VoiceTra began in July 2010. There were 10 million access requests to the VoiceTra at December 2012. The authors also analyze the trends of the huge-scale data obtained in the evaluation.
  Here, although the results of such evaluation are not usually open to the public, the results of the analysis revealed by this study will provide great impact on the developmental direction of the speech translation system.
  As mentioned above, in this paper, the implementation of a practical multilingual speech translation system for mobile devices and detailed analysis of huge-scale evaluations are presented. The contribution of this paper is notable for the future international society; therefore, this paper has been highly recommended for the paper award.

Close