Oki Electric Develops Web-Based Machine Translation System Capable of Enhancement through User Collaboration

Oki will make a thesis presentation about the system at MT Summit VIII, an international conference held every other year by the International Association for Machine Translation.

Oki will also demonstrate a translation portal site named "Yakushite Net" (trademark pending), which means "translation net," at http://yakushite.net/ starting September 18. The site will be evaluated in a trial with the participation of academics and other selected users. During the trial, updated information will be provided to the users via the website.

Oki's new system leverages the cooperative efforts of online users for the creation of highly accurate dictionaries, enabling people with deep knowledge of a particular subject to collaborate in the enhancement of specialized dictionaries for online machine translation. Conventionally, companies and individuals offering machine translation services have maintained their dictionaries on a non-collaborative basis, with only limited success.

"Until now, machine translation has not been used widely because of its low quality. Oki's Yakushite Net, however, will enable internet users to easily improve the quality of machine translations through collaboration," said Toshiki Murata, team leader of Oki's Service Media Laboratory.

Major features of the system (see attachment) are as follows:            (screenshot)

1. Users with a common interest in a specific field can create a "community," or collaborative working group, each of which has a unique domain on the machine translation website. In addition to the system's basic translation dictionary, the community creates and updates a specialized dictionary with words and phrases encountered in its specific field. The community also manages machine translation results, which includes correcting and modifying translations produced by the system. A community can be established for any subject of common interest, such as social sciences, natural sciences or technology, or more specialized categories within these subjects. Communities can be accessed on the website via a directory tree similar to those used in conventional search portals.
   
   
2. Management of community dictionaries is flexible. The creator of a community can serve as the leader and manage the dictionary individually, or all registered members can participate in the management process of adding new words, updating translations, etc.
   
   
3. The system learns new translation patterns based on how users correct and modify the machine's original translation.
   
   
4. Frequently used technical terms are extracted and added on to a community dictionary after given corresponding translations.
   
   
5. Other convenient features include:
   - Glossaries that users can post for translation by other members of the community.
   - Bilingual chatting supported by machine translation.
   - Related links selected by each community.

Two types of dictionaries are available on the site: basic dictionaries that users manage individually and community dictionaries managed on a collaborative basis at the community level. Management of both types of dictionaries includes the ability to add/delete words, perform searches and save user modifications of machine-produced translations (pattern-learning feature mentioned above). Although anyone can use any community's dictionary, only registered members of each community can add entries to the community dictionary.

Notes:

• Java and all trademarks and logos related to Java are trademarks or registered trademarks owned by Sun Microsystems, Inc., in the U.S. and other countries.