In 1994/95, we have begun developing a family of laptop based speech translators, that provide speech translation in a mobile environment.

While the underlying technology is similar to the videophonestation (spontaneous, continuous, speaker-independent translation in limited domains), it now runs on a portable Pentium based device.

Various human factor issues that arise from mobile use have been addressed and solutions proposed.

Presentation of the translation output can be provided acoustically by synthetic speech in the target language, or by a heads-up display that shows subtitles under the face of the other conversant. Here, a heads-up display is mounted on see-through goggles that allow the user to look at another person, while reading translation output on the display.

The portable version of the JANUS system has built-in databases with maps and local information (e.g about hotels and transportation). The user can access this information to find his way through a foreign city. The system is also environmentally aware: using an integrated GPS system to locallize the user and by analizing the background noise, the system knows what level of communication is appropriate in a given situation.

The system can also be used in face-to-face conversations around a conference table, where captions are displayed on the laptop screen.

a conversational speech translator for Spanish.

This situation is encountered when simultaneous translation of speeches, and/or passive translation of conversations between two foreign language speakers are needed.
This type of translation is generally considered to be very difficult for human interpreters as the rapid turn-taking and cross-talk between speech utterances presents highly fragmentary and overlapping input.

In this case the system does not have a record-button nor a send-button. Rather it monitors the two speakers' utterances, segments and processes each of their turns automatically and without human intervention. The translation output is presented for each speaker as printed transcripts. Clicking on the transcribed turns plays back the corresponding portion of the speech signal.

For a general overview you may want to download:postscript "Translation of Conversational Speech" (postscript)

Back to the first page
back to top