Our client asked us:

“We manufacture construction machinery, which is exported into multiple countries in the European Union and Asia. The manuals of the machines need to be translated into the language of every target country. Our previous translation partner sent us the memory from our previous projects. How much (if at all) does this change the price of the manual’s translation?”

More information

This question was raised by a Swedish exporter who manufactures construction machinery. The client sent us the files with the translation memories and the original manuals, and as a result of analysing the translation program we were able to determine how many matches there were. Translation memory is database for sentence pairs, in other words – the heart of translation technology. Its memory consists of previously translated segments (words) and only identical or similar sentences are suggested. The similarity between the source and target segments is indicated with percentages (100% is the perfect match) and lower percentages (0-99%) are called fuzzie matches.

How we solved it

We use a translation memory because it:

  1. …speeds up the translation process – since we already have the previously translated words/segments in the translation memory and we do not need to translate them again;
  2. …improves the quality of the translation, since the same word can be often translated in multiple ways – for example, when translating the English word “book” into Swedish, you can use the term “bok” or “tryckt skrift”. If the client prefers “book” to always be translated as “tryckt skrift”, then the translation memory ensures this is done and the translator does not accidentally translate it as “bok”;
  3. …reduces the price of the translation, because the client has already paid for the matches that come from the translation memory and they do not need to pay for those parts again.

All the above-mentioned reasons are essential, because the client often wants the work done fast and the terminology has to be the same throughout the document. Even more importantly, it allows to us to offer the translations at a cheaper price.

Sample analyses:

TypE Segemnt Word
Perfect Match 0 0
Context Match 140 546
Repetitions 48 196
Cross-file Repetitions 0 0
100% 359 1516
95% – 99% 80 1029
85% – 94% 36 576
75% – 84% 20 252
50 – 74% 11 106
New 86 784
Totalt 780 5005

What means what?

Perfect matches, content matches and repetitions are basically the same as 100% match. This means that a segment in the document is identical to a segment stored in the Translation Memory. In the analyses above, we have 0 perfect match words, 546 context matches, 1516 100% matching words and 196 repetitions. For those the client usually does not pay anything – they have been translated in the past, entered to the translation memory and those are so-called „translated words which the client owns“.

New are new words that need to be translated. In this very case, 784 words.

Percentages, so-called fuzzy matches with gaps 95-99%, 85-94%, 75-84% – these are the words that are partly identical to what is found in TM and need revision. Only 50-74% (sometimes also 0-74%) are the ones where no similar segment has been found and the text must be translated entirely. Those – 0-74% – fuzzies are the ones that confuse clients most often. It may seem that the number of new words is only 784 but in fact 50-74% fuzzies have to be calculated as new ones too.

Do you need a translator?