PRESS BRIEFING ON UNIVERSAL NETWORKING LANGUAGE

Press Briefing

PRESS BRIEFING ON UNIVERSAL NETWORKING LANGUAGE

At a Headquarters press briefing today, representatives from the United Nations University Institute of Advanced Studies demonstrated the use of a computer language -- known as the Universal Networking Language (UNL) -- for multilingual communication on the Internet.

Tarcisio Della Senta, Director of the Institute, reported on the results of a three-day symposium attended by 50 computer and linguistic scientists from around the world. He was joined at the press conference by the Managing Director of Software Translation Artwork Recording, Nikolai Puntikov, and Hiroshi Uchida, the Project Director.

The first results of the research and development were very promising, Mr. Senta said. The Language could bring enormous benefits for education, communication and "E-commerce". However, more research and development was needed and it would take many years to include all languages.

Mr. Senta said the new language would enable people to communicate with each other in their different native tongues by using the computer as an intermediary. Because of the language barrier, many people around the world had less access than others to knowledge and information on the Internet.

Mr. Puntikov said UNL was a computer language which would be understood and interpreted by a computer programme. It was a technology -- or broad network of language servers around the world -- which converted language texts into UNL language. The language was an example of truly international cooperation between the best research groups and businesses working in language engineering. To date, 17 groups were working in 17 countries, but by the end of the project in the year 2006, 150 languages would be covered.

The Universal Networking Language was not another machine translation project, he continued. The goal was to create an interactive human computer mechanism which would allow people who had no specialized knowledge to work with UNL. It paved the way for truly global communication. Mr. Puntikov then gave a visual demonstration of how UNL worked.

The Universal Networking Language consisted of three interlinked computer software programmes residing in the World Wide Web, accessible to any Internet user and compatible with standard network servers, he said. Text written in a range of languages might be converted into UNL ("enconverted") and, just as easily, UNL text could be converted into native languages ("deconverted"). Unlike traditional language conversion methods, UNL avoided language analysis by utilizing the common platform provided by the Internet.

For example, the text file of a press release in Hindi would be sent to the Hindi language server that would help create the UNL representation of the

UNL Press Briefing - 2 - 23 November 1998

meaning of the text, he said. The text would then go to language servers corresponding to the target audience.

Scientists had been working on the project for two years, he continued. At the symposium, it had been decided that enough progress had been made to promote the Language to the wide community of Internet servers. It would be published in the open domain so that everyone would have access, and trial unit servers would be established so that the technology could be tested.

Mr. Uchida said the 10-year project had started in April 1996, and currently there was a United Nations system development kit. Sixteen language models were being developed. By the year 2000, the project aimed to have a practical service for Arabic, Chinese, English, French, Russian, Spanish, German, Hindi, Indonesian, Italian, Japanese and Portuguese.

A correspondent asked how long it would be before the system would be used. Mr. Senta said the Language was currently only available to the network of scientists and researchers working to help improve the technology and quality of the service.

In response to a question about whether it had been tried out with real English text, Mr. Puntikov said that this year, most of the research groups would publish on their web servers, pages that could be put into UNL text. They would be able to check the results of the conversion in major languages.

Was it possible to translate nuances, or was the system restricted to the language of mathematics and science? a correspondent asked. Mr. Senta said that in the early stages, UNL was not intended for translating poetry or Shakespeare. It was more suited to the language of mathematics, science, E-commerce or software manuals.

In response to a query about whether the UNL would encounter the same problems encountered in the field of machine translation, Mr. Puntikov said it was not another machine translation system; rather, it involved analysis by a human/computer interactive mechanism. People would resolve ambiguities in a formal dialogue with a computer -- it was not mere translation of texts.

In the future would it be possible to codify nuances? a correspondent asked. Mr. Uchida said he did not think so. The initial aim was to convey information.

* *** *