The localization industry today is in a state of change. The recent evolution of cloud technologies has finally given us what we’ve always dreamed about as translation professionals: safe, secure storage of translation memories in a single location accessible by the entire world. Add an easily programmable search routine that everyone is familiar with (think “Google”) along with a user-friendly interface and the industry can finally shed the decades-old, desktop-based systems that have held this industry captive. But where to start and how to integrate? How do we break old traditions and introduce a new generation to new tools of the trade? We need to start somewhere, and I suggest we begin with the elimination of “fuzzy” translations, an ancient concept that has lost its place.
A “fuzzy” translation is one that almost matches an existing database translation, differing by some sort of pre-determined percentage. Often this can be just a number, such as when comparing “Go to step 1” to “Go to step 2”, or it can be a single or multiple term phrase, such as “When you arrive, lock the car” compared to “After you park, lock the car”. Translation tools are programmed to determine some level of “fuzzy” matching in existing translation databases that can then be leveraged for the project. Agencies offer a discount to clients for this leveraged work, and everyone (agency and client) feels like they are getting a great deal.
It does sound like a solid savings, right? Instead, I believe it’s misleading, and the term “fuzzy” should be eliminated from our industry.
Here’s the flaw. Translation engines have always been based on the “segment” which is typically a sentence, but could also be a fragmented sentence, title, column heading, etc. depending on its use in the document. The engine generally displays the “segment” to the translator who produces a translation that is then deposited in the translation memory database.
Up Next: The “tools” light the way
