"Raw data, including translation memory data, has no value per se".
Or so is the opinion of Jost Zetzsche, as expressed in the current Translation Journal. In the article, he ponders the sudden availability of large amounts of bilingual data that can be used in translation memories ("TMs").
What is interesting about Jost's piece is that it goes beyond the question of "is a huge, shared TM a good idea?" (we have weighed-in on this before) and poses a more fundamental question:
Is bigger really better when it comes to TMs? Or: Is there a point where a TM jumps the shark by becoming too large, too unwieldy, and too cumbersome?
Jost does a nice job providing pros and cons. And from our own experience here at ForeignExchange, I can also say that it's impossible to give a black-and-white answer. However, what is very clear is that size for the sake of size is not a good thing.
In fact, we have recently gone the other way, proactively reducing the size of TMs by separating them (by client, division, product group, etc.) and culling segments based on age or attributes. Like Jost, we have started to wonder whether or not we are spending too much time updating old, outdated segments.
This "small is beautiful" approach is yielding better output and an overall lower cost - even though we are pulling somewhat fewer matches from TMs. Like all TM efforts, though, it needs to be managed.
The appeal of large "Big Mama" TMs is that they're easier to administer. Just throw anything in there. Being selective about TM contents means that a human specialist has to make manual determinations about what segments should and should not be included. In the current economic environment, that flies in the face of "doing more with less".
What is your approach to TMs: Small and beautiful? Or Big Mama style?
ForeignExchange Translations provides specialized medical translation and software localization services to medical device and pharmaceutical companies. Contact us to learn more.
Categories: translation memory