"ChaTra" the speech translation application in smartphone: real-time multilingual communication by multiple people

October 13, 2011

The National Institute of Information and Communications Technology (NICT, President: Dr. Hideo Miyahara) has successfully developed “ChaTra”, a speech translation application by which multiple people can communicate using a smartphone like ‘chatting’. This application is originated from the so-called “VoiceTra”, which was released to the public in August 2010, and has been developed so that users can make use of communication by connecting their own terminals (either iPhone or Android).

Background

With the aim of overcoming the language barrier, NICT has been addressing the research and development of machine-based multilingual speech translation technology and challenging the goal of “Achievement of speech communication technologies to overcome the language barrier”, one of the “Projects of accelerating contribution to society” driven by the Japanese government. NICT started the field experiments of a multilingual speech-to-speech translation technology, by releasing the free-service, multilingual speech translation applications, ‘VoiceTra for iPhone’ in August 2010, and ‘VoiceTra for Android’ in April 2011. These applications have won the great popularity among users, reaching the record hit of more than 500, 000 downloads since their releases.

It is under such circumstances that we have been advancing the study on the usage of systems for multiple users’ conversations via their individual terminals, as well as for peer-to-peer communication via a single terminal.

New Achievements

Screen of Start-up

NICT has developed “ChaTra” which has aided us in realizing conversation enabled by multiple people using their individual terminal devices such as iPhone and Android. “ChaTra” is originated from “VoiceTra”, which is equipped with a user-friendly interface, covers 21 languages with text input/output (6 languages of them with speech input/output) and performs highly accurate translation for the travel domain. With the realization of this technology, it has now become possible for an increasing number of users to connect and make conversation through speech translation on a global scale, regardless of device models and telecommunications carriers.

“ChaTra” is a network-based speech-to-speech translation application which is facilitated with the protocol standardized in ITU-T, making it possible for people living anywhere in the world to connect to processing servers for Speech Recognition, Machine Translation, and Speech Synthesis, using the internet via a 3G or Wi-Fi connection (Telecommunication charge is required separately depending on the telecommunications carrier). The maximum number of users who can use the “ChaTra” service at one time is limited to 5 people. Within 2011, NICT is scheduled to make this application available to the public for free in order to foster the feasibility and usability study in the real fields and bring about problem analysis and technological improvement.

Future Plans

NICT aims to create business opportunities along with the field experiments in cooperation with private businesses, and to make further contributions to society by leveraging our research results. NICT also has the future plans for field experiments of speech translation in collaboration with some institutes all over the world in the next fiscal year.

*	VoiceTra is a registered trademark of NICT.
*	Android is a trademark of Google Inc.
*	iPhone is a trademark of Apple Inc. and registered in the U.S. and other countries.

Supplementary information

1. Image of using “ChaTra”

1)	Start the application by tapping the icon (Fig.1) for “ChaTra” installed on the smartphone.
2)	After the application has started (Fig.2), select the language to use in the setting screen (Fig.3), and tap ‘Complete’.
3)	Start chatting by nominating the users you want to invite in the registered list.

Fig.1 Icon

Fig.2 Screen of Start-up

Fig.3 Screen of Setting

Once Chatting has started, the results of chats in the selected language can be confirmed in the chat log.
Fig.4 shows screens for a case in which users are chatting using the selected languages of English and Japanese. English is shown to the user who selects English, while Japanese is shown to the one who selects Japanese.

Fig.4 Screen of Chat

*	“ChaTra” covers translation with text input/output for 21 languages, including speech input/output for 6 languages.
*	When starting a chat, select the users who wish to invite.
*	Registration of users is required in advance to identify the users.
*	The maximum number of users who can be enrolled for one chat is 5 people, and the time for chat is limited to one hour.

2. ITU-T Standards

Fig.5 Documents of Standardization

“ChaTra”’ applies network-based speech-to-speech translation based on a protocol (H.645, F.745) for which standardization was realized by NICT at ITU-T.