What is a fuzzy match?
What are 100% and fuzzy matches – and how would we charge for it?
We all know duplicates – which are sentences which appear more than once within a text. If a translator has a CAT tool, she only has to translate duplicates once and it will be “propagated” through out the text.
But if you have a Translation Memory (TM), then you might find that some sentences in your new text have already been translated. This is where “matches” come in.
What’s a 100% match:
Translation Memories store source sentences together with the translation. If the text on which the TM is being applied contains a sentence that is identical to one in the TM, then we have a 100% (or “exact”) match.
The “fuzzy match”
If a sentence in the new text is not identical, but “just” similar to one in your TM, then you may have a “fuzzy match” - consider this example: your TM contains the sentence (including the translation, of course):
“Everybody in Auckland’s traffic jams listens to the radio”
The new text contains this sentence:
“Everybody in London’s traffic jams listens to the radio”
In this case a CAT tool will most likely flag the new sentence and offer the first sentence’s translation as a “fuzzy match”. The translator can then decide if the context is ok and change what has to be changed (in this case the city name only).
Obviously in this example we have a high fuzzy match that hardly requires any work by the translator. It may be reasonable to give the client a discount for such fuzzy matches. But not every fuzzy match has the same value to the translator. Consider, for example, this sentence:
Everybody listens to radio broadcasts about London’s traffic jams.
This sentence may be a fuzzy match, but there is much more work required to edit the translation in the TM to get a good translation.
Even a 100% match too has to be carefully checked by the translator, because there may be errors in the translation, or the context of the sentence is so different that it needs to be re-translated.
Obviously you only get fuzzy or 100% matches if you have a TM – and you need a tool to analyse the new source text “against” the TM.