Translation and Interpreting in 150+ Languages
Translation Memory Ninjas: Over the Wall
September 17, 2012 - By: - In: Translation - 13 comments

Indeed, the three treasures of translation are the translation memory, the terminology base and the style guide. Without the three treasures, translation is chaos.

First among these treasures is the translation memory. Those who follow the righteous path of translation memory will prosper. Those who do not are doomed to the endless cycle of retranslation and the repetition of error.

-The Art of Translation 

The wise translator heeds the ancient worthies. But not every translator is wise. For a technology so critical to translation best practice, the scope of translation memory abuse in the translation industry is a constant surprise and delight to me. Profit is sweet, and profit from the carelessness and ignorance of my competitors doubly so. Identifying and correcting the translation memory errors of our competitors is a dramatic and surefire way to win new business. Was it not Sun Tzu who said that it is better to capture the enemy’s wagons than to burn them? I think so. Something about oxen, I’m pretty sure.

So. When viewed as a wagon, translation memory (TM) seems simple enough.
1. Pair a phrase with its translated counterpart.

2. Reuse the same translation every time the original phrase appears.

3. Repeat.
Our un-wise competitors do a file dump and use a TM tool like Trados to spit out a count of unique, repeated and fuzzy-matched phrases. That was easy! Copy and paste the analysis into the quote, editing the numbers to suit profit goals, and you are on your way to the bank. Less repetition, more money. Unless your carelessness is discovered by the master (dramatized in the photo above) of our translation memory ninja house perched high in the misty mountain fastness of Iga. (Never-before-seen footage of this top-secret facility revealed here for the first time.)

As soon as our TM ninjas get hold of  a memory or source, they slice and dice it like so much sashimi. We actually look at look at structure and content of source first to figure out how to increase repetition! Very sneaky!

A few years ago, we got wind of one of those juicy six-figure engagements, a million words plus. Lionbridge had tossed the files into their TM sausage grinder and came up with a quarter-million-dollar quote and 3% repetition. Great, except the prospect didn’t want to pay a quarter-million dollars (Clients are like that, I’ve found.) So when they complained to us, we sent our ninjas over the wall. First thing, they ditched Trados for a tool that allowed them to parse the segments more tightly, then our ninja coders wrote a few lines  to pull a lot of extraneous text out of the TM process, along with some other secret TM ninja tricks that I fear to disclose. By actually paying attention to the way the source file was constructed, our ninjas brought the job in at 7% our competitor’s estimate, and we won a Fortune 100 account to boot. I would have liked something a little closer to six figures, but while our ninjas are extra-sneaky, they don’t cheat on word counts. Granted, this was an exceptional situation due to the unusual nature of this project, but controlling source is the secret sauce in any translation automation technology, and the sweet spot for the dramatic service and cost differentiators that win client hearts and wallets.

The moral of the story: Translation memory is harder than it looks.

Next time, translation memory ninjas defend the palace. Until then keep your TM shuriken sharp.

LiveZilla Live Chat Software