Machine Translation Systems

Since the earliest days of computing, automatic machine translation of natural human languages has been a dream. As a result, there are a great many machine translation (MT) systems. Because the leading edge of computer technology was in English-speaking countries, most machine translators include English in the language pairs they support. With the globalization of technology, and with the exponentially increasing power of computing, we expect that machine translators will soon appear in all locales.

Considering the "gift" model of the World Wide Web and its spirit of collaborative development, we expect that most machine translators will be freely available over the web. Some will no doubt charge for an enhanced level of translation services. Many will be supported by advertising revenue. But whether commercial or free, in order to interoperate easily, they will all need to agree upon interface, database, and data interchange standards.

The centers of machine translation development have been universities on the one hand and leading computer companies on the other, with the companies often funding the academic research. Commercialization of MT systems ranges from desktop translators that cost under $100 to enterprise-scale translation incorporated in content management systems that cost millions of dollars.

Babel, a joint I18N project between The Internet Society and Alis Technologies, is an ambitious effort to allow a web browser to work in any language with just-in-time translations and character code conversions (to the 16-bit unicode needed by non-Western languages). Alis "Gist-in-time®" is incorporated in the Netscape 6 browser, and their other components make the Windows OS multilingual.

  Babelfish (now babelfish.altavista.com) is the best known free (advertising-supported) machine translation service on the web. It provides immediate online translations from French or English into 5 European languages and back. You can test their translations of some basic web phrases here. The underlying translation technology is by Systran. You will find Systran translation technology on many web sites, like Dictionary.com and Lycos. It can be purchased from Systran for use on your web site.

Transparent Language also offers immediate online translations using their TranscendRT technology at freetranslation.com. TranscendRT and freetranslation.com were recently purchased by SDL International, who now offer online machine translation on their site.

InterTran from TranslationExperts, Ltd. offers a few dozen language pairs, and an interesting sentence diagram feature. Optional translations are suggested for key words and words can be rearranged on screen.

Reverso and PROjectMT, Ltd add Russian to the FGS language pairs. PROMT Reverso technology is used on many sites world wide, Softissimo and VOILA in France, Smart Link in the USA, and PROMT themselves in Russia.

Perhaps the widest ranging online MT portal site, with access to 22 machine translation sites working in more than 50 languages is foreignword.com. They also offer links to 178 online dictionaries, 1001 glossaries, and hundreds of translators.

Powered by

Most of the free web translation sites also let you enter a URL and they will translate a complete web page for you. They all preserve the web page HTML. Some even substitute links to themselves for all the hyperlinks in the page. This means you can navigate links on the translated page, and they come back translated. This can be a benefit and a hindrance to web developers looking for gist translations. Many millions of web pages have been translated by all these free online services. They will play a big role in the future of the multilingual Web.

At the other end of the spectrum from free web machine translations, large corporations are buying huge "enterprise translation servers." These will be centrally located and provide over-the-Internet translations to their users from any browser. Large translation companies hope to sell real-time web-based translation services, either by translating web pages on-the-fly as users request them, or by returning translated web pages to the smaller company's web server. Typical rates are pennies per word for the raw machine translations, and $.25 per word for human-corrected text. These companies have armies (thousands) of translators, who will be able to improve the gist machine translation and return a human translation to the server (or to a client) with very fast turnaround times. (Note that many translators proficient in the source and target languages find the bland computer-speak gist a waste of their time). The entire translation business - quotations, scheduling, translations, approvals, and billing, will be conducted over the web. Translation suppliers may never meet their customers. Translation will be a universal web application, working on web application servers, at application service providers (ASPs).

Systran Enterprise costs about $5000 for a single language pair and five users, $32,000 for eight language pairs and 20 users. Lernout&Hauspie say their iTranslator Enterprise is coming soon. The Transparent Language MT engine called TranscendRT has been acquired by SDL International. Their Enterprise Translation Server is priced starting at about $17,000. The IBM WebSphere Translation Server offers translations at 500 words per second from English to FIGS and CJK languages for $10,000 per language pair per CPU. Wordstream is developing a multilingual translator for the Internet called ClearText. It can translate a stream of text (like a news feed) into multiple languages in real time. No prices have been announced.

A pioneer of machine translation software, Logos, has been acquired by globalwords of Germany. Globalwords is primarily a telephone-based real-time voice translation service. A similar organization is Free Translation 1-800-Translate, offering translation, interpretation, and localization in 148 languages, plus several machine translation gists.

All these efforts at machine translation (MT) have received mixed reviews. But see also the Atlantic Monthly article, "What Global Language?" Leading companies that have invested heavily in MT have done badly as the Internet frenzy has cooled and dot-com investments are on hold by many venture capitalists. Over a billion dollars has been invested in machine translation, mostly by the U.S. Defense Department and CIA. Much work has been done in universities. Carnegie Mellon University is a recognized leader in MT. The European Commission Systran MT (a distant relative of U.S. Systran) machine translates about a million pages a year into its 11 official languages. Apart from those with large government contracts, and science-fiction fans looking for Douglas Adams' Babelfish or Star Trek fans after a Universal Translator, knowledgeable observers doubt that machine translation will ever translate the subtle nuances in everyday language. Machines are now seen by language professionals as aids to humans for Computer-Aided Translations (CAT). Machines can provide the gist of a document for someone who does not know the source language.

Machine Translation Systems On The Web

Most of these systems will translate text or a URL. Be aware that they are somewhat fragile sites, even more so with all the flux in the globalization dot-com marketplace (not to be confused with www.globalization.com, a web site for SDL International).

Alis Technologies Gist-In-Time English, French, German, Italian, Portuguese, Spanish, Japanese
babelfish.altavista.com English, French, German, Italian, Portuguese, Spanish
Ectaco Albanian, Arabic, Bulgarian, Czech, French, German, Greek, Hungarian, Italian, Korean, Latvian, Farsi, Polish, Portuguese, Romanian, Russian, Serbo-Croatian, Slovak, Spanish, Swedish, Turkish, Yiddish
e-lingo English, French, German, Italian, Portuguese, Spanish
freetranslation.com English, French, German, Italian, Spanish, Portuguese, Norwegian
GPLTrans Dutch, English, French, Indonesian, German, Italian, Latin, Portuguese, Spanish
IBM alphaWorks and IBM developerWorks English, French, German, Italian, Spanish, Japanese, Chinese (S), Chinese (T)
InterTran Bulgarian, Croat, Czech, Danish, Dutch, Filipino (Tagalog), French, Finnish, German, Greek, Hungarian, Icelandic, Italian, Japanese, Norwegian, Polish, Portuguese, Portuguese (Brazil), Romanian, Russian, Serbian, Slovak, Slovenian, Spanish, Swedish, Welsh
Language Engineering Corporation LogoVista X English, Japanese
Lingvistica '98 English, German, Russian, Ukrainian
Poltran.com (Ectaco) English, Polish
PROMT English, French, German, Spanish, Russian
Reverso English, French, German, Spanish, Russian
Russian Translation USA English, Russian, Ukrainian
Rustran.com (Ectaco) English, Russian
Systran English, French, German, Italian, Spanish, Portuguese
Transbot English, French, German, Spanish
WorldLingo English, French, German, Spanish, Dutch, Russian, Chinese (Simplified and Traditional), Japanese, Korean
Universidad de Oviedo. English, French, German, Spanish

 
Criticisms? Suggestions?   Do you want us to add a reference or hyperlink to this page? Send us an email or add a comment directly to this page below. Back to top of OIL site.

Altavista Babelfish

Content Management Professionals

CMS Review

GlossPost

Information Architecture Institute

Information Philosopher

Interaction Design Association

Memetic Web

ProZ - Translators Workplace

skyBuilders

TaxoTips

 workFlow  |  Subscribe