Just as the U.S. government and international relief agencies must compensate for a dearth of human translators, global businesses may be able to use software to help workers speak-or least hear and read-each other’s language. Language translation software at least should help workers get the gist of a policy or product update, in their native tongue. However, “machine translation” software that autonomously translates any given text from one language to another can be error-prone.
Problem: Machine translations can produce text that is garbled or hilariously inaccurate.
Resolution: Test the precision of your translated text by sending a phrase on a round trip through the translation engine. For example, a sentence that appeared recently on the Baseline Web page-“Why should we rely for patches on the company that made the holes in the first place?”-turns into something about a “lean piece of cloth” when translated into Korean and back by the Babelfish.altavista.com Web site, which uses software from Systran, a French machine translation firm. Simplify; make more straightforward; try again.
Problem: Basic translation software often fails to pick up on intended connotations. For example, the word “chip” can mean entirely different things to semiconductor researchers and snack food distributors.
Resolution: Find translation software packages tailored to support your industry or topic domain. Systran’s Enterprise and Premium Language Translation Software packages include 20 industry-specific dictionaries. When the computer dictionary is turned on, for example, the word “server” will be translated into a word that means computer server, not waiter or waitress.
Problem: Even industry-specific translation software can fail on sentences that use idioms, slang or jargon that the software hasn’t been trained to recognize.
Resolution: Make sure writers and editors use simple declarative sentences and a limited vocabulary when creating a manual or other document that will be translated into many languages. Translation software produces better results when it handles short, simple sentences.
Problem: Machine translation is considered too bleeding-edge for many business applications, such as providing financial, supply chain or shopping services to overseas customers.
Resolution: Look into workflow and content management software that can assist human translators.For example, some content management software packages, such as those from Trados, feature “translation memory,” which stores equivalent blocks of text in multiple languages. The translation between languages is accomplished by humans, not software, but these reusable chunks of content can be called up and used when, for example, a customer service representative is going through a brochure, warranty or other business document with a customer.