| You are here: | About>Education>German Language |
 | German Language |
|
A Perfect German Translation?Is There Such a Thing?Why are computers such lousy translators?The Problems of Machine Translation In our previous article about translation software, we pointed out the deficiencies of machine translation today and the obstacles facing developers trying to improve computer translation programs. The May 2000 issue of Wired magazine focused on the current status of machine translation (MT) and artificial intelligence (AI). In a series of six articles under the rubric "Hello, World," the popular Web magazine examined the present and future of translation in great detail. "Imagine a machine that speaks your language — and translates it for those who don't," reads one heading in Wired. It's an attractive prospect for the average monolingual US-Amerikaner (the German term for people from the US), but just how likely is this dream of the universal translator à la "Star Trek"? Well, Wired seems to agree with my own assessment of the state of the art of MT. The articles point out just how impractical current translation software is for any serious use, and how rarely it can be relied on — without some human intervention and editing. But the Wired articles also point to the future and some interesting developments in both AI and MT.
“There is no such thing as 'perfect' translation. There are only translations more or less suitable or successful for specific purposes and contexts.” - MT historian John Hutchins (in Wired) The Wired piece includes a map of various "hubs for machine translation R&D worldwide" — including the DFKI, or Deutsches Forschungszentrum für Künstliche Intelligenz GmbH (German Research Center for Artificial Intelligence GmbH) in Kaiserslautern and Saarbrücken, Germany. Founded in 1988, the non-profit DFKI heads up a consortium of over 20 universities and corporations for several AI/MT projects funded by the German government and German industry. DFKI's partnership includes large German companies such as DaimlerChrysler AG, SAP AG and Alcatel SEL AG, as well as many smaller firms. The Verbmobil project ("Multilinguale Verarbeitung von Spontansprache") is using AI in an attempt to develop a special program for translating from or to German and English or Japanese. Verbmobil has the specific goal of producing and translating spoken language that sounds natural, complete with intonation and other human qualities, for the travel industry. (Think, hiring fewer human telephone agents.) Code-named "Janus," DFKI's German translator is already in a demonstration stage that can translate spoken English into audible German (or vice versa). Janus is capable of ignoring our very human "ah's" and "um's" and translating the essence of a spoken utterance. But Janus still has serious limitations. When a tester intentionally strays from normal "travel" vocabulary, "astrophysics" is transformed rather oddly into "Mastercard." This serves to illustrate a maxim from MT historian John Hutchins quoted in the Wired article “Talking to Strangers.” In explaining that even infinite computer power is not the answer to better machine translation, Hutchins says: “There is no such thing as 'perfect' translation. There are only translations more or less suitable or successful for specific purposes and contexts.” In the same article, Douglas Hofstadter, author of the book Le ton beau de Marot, bases his criticism of MT on what he views as its fatal flaw. Speaking of the many examples of error-filled computer translations, Hofstadter points out: "Programmers will say, 'That's the kind of thing we can polish up very quickly.' Or people will say, 'That's because the programmers were bad.' But the most ordinary vanilla language contains a tremendous amount of subtle knowledge — what space is like to move around in, how people are, what people want. We know the structure of the space because we've lived there." He believes that a computer program is unable to duplicate these subtleties or to mimic the human brain's ability to manipulate language to suit a variety of human situations. That's why many researchers and developers claim that solving the MT problem means solving the AI problem. They claim that a more complete form of artificial intelligence is essential for anything approaching good machine translation, and that the current semi-mechanical approaches to MT are doomed to failure. To be a good translator, they maintain, a computer program needs to "learn" the languages and gain some knowledge of the real world – just as we humans do. The bottom line in all of this is that the more we study language and translation algorithms, the more we realize how truly complex human language is. And that explains something else important to language learners. If language were simply a bunch of words strung together, it would be a lot easier to learn another language. As it is, those of us trying to learn German as a second language can take solace in the fact that even a high-powered computer finds human language "a tough nut to crack" ("eine harte Nuss").
MORE > German Translation - Articles & Resources
Related Pages
Translation Software
Interview 1: Translating Karl May
Interview 2: Translating Winnetou
Shakespeare in German
The Bible in German
Translation Page
On the Web
DFKI - German Research Center for Artificial Intelligence
DFKI Newsletter
"Talking to Strangers" - Wired
United Nations - German Section
More Related PagesMORE > Free German Translation
German Newsletters German for Beginners
|
|
All Topics | Email Article | |  |  |
| Advertising Info | News & Events | Work at About | SiteMap | Reprints | Help | Our Story | Be a Guide |
| User Agreement | Ethics Policy | Patent Info. | Privacy Policy | ©2008 About, Inc., A part of The New York Times Company. All rights reserved. |