Translate web-based info

The World Wide Web connects computers from around the planet for sharing information. With web technology, a person in one corner of the world can access a page hosted on a server on the opposite side of the globe. But will that person be able to read the page?While the web is global in scale, most of its content is in English.

By Matt Bellm July 1, 1998

The World Wide Web connects computers from around the planet for sharing information. With web technology, a person in one corner of the world can access a page hosted on a server on the opposite side of the globe. But will that person be able to read the page?

While the web is global in scale, most of its content is in English. The Internet Society (Reston, Va.) and Alis Technologies (Montreal, Quebec, Canada) teamed up for the Babel project to look at language issues on the web ( www.isoc.org:8080/ palmares.html ). Preliminary results of the Babel study shows that 82.3% of web pages are in English (see table).

Euro-Marketing Associates, a San Francisco-based international marketing agency, reports that only 28% of Europe understands English. This means that most of Europe, let alone the rest of the world, cannot understand most of what is on the web. France has taken an aggressive position to overcome this imbalance. A 1994 French law states that any good or service sold in France must be solely in French or in French with a translation to other languages. This has been interpreted to apply to web sites hosted in France. While France’s efforts to promote its own language within its borders will help increase the number of pages French-speaking web surfers will find in their native tongue, it doesn’t ultimately help make the web-based information more accessible to the world. To make sites more accessible web developers could translate pages into multilingual versions. However, development and hosting costs for an expanded site would increase.

Removing barriers

Ability to quickly and accurately translate web pages would help tear down some of the web’s linguistic barriers. Software for translating web pages is available from numerous companies.

One example is free for anyone to try. The AltaVista search engine (altavista.digital.com) provides a page called Babelfish for on-the-fly, machine (automated) translations. A button on Alta Vista’s search pages leads to the Babelfish, which uses Systran ( www.systransoft.com ) translation software. A text field on the Babelfish page lets you type text to be translated or the URL for a web page to be translated. The page translates to and from English, French, German, Italian, Spanish, and Portuguese.

When a URL is entered, Babelfish translates the text from that web page and places it back in the page with its original graphics and layout. It does not translate text contained in graphics, so graphic navigation buttons would remain in their original language. Babelfish translations provide a good idea of what the page contains, but machine translations are not always intelligent enough to understand a word’s context. For example, the results of a French-to-English translation contained the term “web ferreter” when the correct term would have been “web browser.” At this writing, AltaVista labels Babelfish as a beta version.A new development in the translation market from Lernout & Hauspie (Burlington, Mass.) provides an integrated system for accessing web-based information. The system, called Coronado, not only translates web content, but it also helps users find relevant web pages regardless of language.

Users specify search terms and which search engines to use with the Coronado client software. To add context for more accurate searching, users can also specify content areas, such as engineering and computers. User search criteria is translated to supported languages and submitted simultaneously to multiple search engines in multiple languages. When the client displays search results, summaries appear both translated and in their original language. When a link from the search results is selected, the page is translated on-the-fly at the server and displayed in a browser with its original graphics and page layout. After closer inspection, if a user wants a more accurate translation, pages can be submitted for human translation. Human translation is paid for on a per page basis in incremental blocks.

Translate on demand

As an alternative to translating a web site into multiple languages, Lernout & Hauspie plans to sell their server-side translation product to individual companies. Companies could then provide multi-lingual versions of their web site without creating multiple versions of web pages. Users could specify the desired language and pages would be translated as they are selected. Lernout & Hauspie products currently translate to and from English, German, and Spanish. French will be added in mid-1998 and Japanese, Chinese, Dutch, and Portuguese are in development.

Author Information

Matt Bellm, Internet Editor, mbellm@cahners.com