Where do we go to sign up? – John Goff
Get Your Adverbs Here
While off-the-rack translation software still leaves a lot to be desired, proprietary programs are getting better. Typically, applications employ complex semantic rules to perform syntactical and grammatical analyses on text. Such programs identify nouns, verbs, adverbs, and adjectives within a sentence, then access a database or dictionary for the correct translations. Finally, the software rearranges the parts of speech into the correct sequence.
By contrast, translation memory software — mostly an aid to human translators — delves into a database for perfect matches in previously translated sentences. The technology is less flexible, since results are affected by such minutiae as spaces between words. But recent developments in “fuzzy matching” allow some translation memory systems to recognize similar, not just identical, matches.
Currently, there are tons of browser-enabled translation programs on the market. One of the more popular is operated by Paris-based Systran (www.systransoft.com). Initially, Systran designed information translation and retrieval software used primarily by the FBI, CIA, and European Commission. But in 1997, Systran CEO Dimitrios Sabatakakis got AltaVista to incorporate Systran’s translation engine into its Net search portal. Now Systran’s technology is linked to more than 300,000 Web sites, from small dotcoms to Fortune 500 companies.
Essentially, Systran functions on an ASP model. Fees depend on the scale of customization and the number of users. “The big advantage of machine translation,” claims Sabatakakis, “is that you constantly add terms to the dictionary, so you build a resource for the company — one that can’t quit its job if it wanted to.”
Brussels-based Jonathan Sage and Dominic Kelleher, PricewaterhouseCoopers’ (www.pwcgloba l.com/us/) directors of knowledge management for Europe, the Middle East, and Africa, are proponents of the technology — to a point. Two years ago, Sage says, business leaders at the consultancy created a Language Action Committee because half the company’s workers are not native English speakers. Moreover, Sage says, the Internet is creating a new demand for speedy multilingual communication.
After analyzing machine translation programs, the committee settled on software from Systran. In December, after six months of integration and customization, the Systran translation facility was added to the consultancy’s corporate intranet. Sage says PwC’s professional internal translators use translation memory software from vendor Trados (www.trados.com), as well.
Now that PwC employees have their own translation home page, managing user expectations is key, notes Sage. Toward that, PwC has included a “health warning” on its Web translation page, advising staff about appropriate usage. Workers are encouraged to use the technology when they want to get the gist of a document’s meaning — say, tax guidelines published in a local language. No-go zones? “Any translated communication or document that’s headed for clients should always pass through human translators,” Sage says. “The same for key internal communications. Otherwise, it could get farcical.” – AL