Since the days of the Phoenicians, trade has depended upon human translators and interpreters, people who have mastered at least two languages and who have acquired the skill of moving ideas from one language to another. But while old-fashioned clerks were mechanized away by the word processor, old-fashioned translators are still with us.
Now that your work station has more computing power than the average research university did twenty five years ago, it’s logical to ask: can’t all that computing power be put to work? If you’ve been reading the ads in trade publications, you know that there are plenty of applications that promise to do just that.
When it comes to spreadsheets, it’s easy to calculate the cost effectiveness of software. But when it comes to language, which is precisely what sets humans apart from the rest of the animal world, the goal may be impossible. Every word the Europeans say to each other within the context of the European Community has to be translated- a huge requirement. With access to the most advanced machine translation methodology available, the EEC has found that current technology is suitable for only about five percent of their total translation volume.
Multinational corporations have also spent many millions working on the problem. In the process, this fundamental dilemma has become crystal clear: language still baffles formal analysis.
There are, of course, achievements to report. Some companies, including Xerox, use MT (machine translation) for their manuals, with the original English versions written by controlled language specialists, and the output edited in-house by qualified professional translators.
While they save with economies of scale, this is still an expensive process. We recently estimated the cost of producing just the Spanish glossary for GM’s CASL (Controlled Automotive Service Language) at $120,000.
How about all those programs you see advertised? That depends on what you expect from them. Most are based on a dictionary of several thousand words and stock phrases in the target language that substitute for words and phrases you enter in the source language. If you feed the computer a whole sentence, it looks for the largest corresponding unit phrase in its memory, then handles the remainder essentially word-for-word.
Much of the time this yields a product that is at least comprehensible in the target language. But be warned: results often fly wide of even that modest mark. One popular Spanish program does fine with the sentence “I’ll arrive in Madrid on the first of May”, but its rendition of “I’ll arrive in Madrid on May first” makes sense only if your horse’s name is “May First”.
That’s why many of these programs are gathering dust in the corners of American offices. They work fine- as long as you don’t demand consistent accuracy and unambiguous clarity. Our translators use MAT (machine assisted translation). It’s a technology that holds your manual translation in memory, and leverages it for revision management, while it creates and expands a glossary. In skilled hands, it will save you time and money.
We are accurate. We are on time. You can depend on us, and we trust you to recognize the value in that. Our clients do. Just ask for a reference.