What is Relaxed Tag Parity and how to best enforce it by changing Studio penalties in tm.properties? We tell you about it in an article!

WorldServer users might wonder why they are getting a 99% fuzzy match from their Translation Memory for a specific segment they have translated before instead getting a 100% or ICE Match. Looking more closely, you might notice that in previous translation, the translator has added or removed some formatting tags in the target segment while working in Trados Studio. As a result, there is a tag discrepancy between source and target segment. The tag discrepancy might have happened by mistake. In fact, this is a quite common mistake. We strongly recommend to always run a Tag Verification in Trados Studio and fix any tag issues before creating a Return package to be uploaded back to WorldServer.

However, sometimes the tag discrepancy is not the result of a mistake, but it is there for linguistic reasons. For example: the word "4th" or "3rd" etc. as an ordinal number needs a superscript in some languages but not in others. For example, you might have an english source segment like:

This is my 4th visit in this city.

The italian translation would require a superscript character which can be easily inserted in Studio:

If you change the view in Studio, you will see that the superscript character is associated with Tags:

These tags are now added to the target segment, but they are not present in the source segment. Once your translation has been imported in the WorldServer TM, the associated TM entry will reflect the translation in Studio. It will include the additional tags in the target segment that are not present in the source segment in the TM entry. As a consequence, if the same or a similar source file that includes the same source segment is resubmitted later using the same language combination and Translation Memory, during segmentation WorldServer will find this TM entry but it will apply a penalty to it, leading to a 99% Match instead of a 100% or a ICE Match for that segment. The name of the penalty applied in this case is the Different Formatting Penalty. This penalty is configured by default in WorldServer. This is just an example of how a formatting issue conflicts with a linguistic requirement. So what can be done about this? How do you get a full TM leverage (100% Match) in this case?

This article provides the answer. It explains the concept of Tag Parity and how to enforce it by removing the formatting penalties, if appropriate for your projects. It also warns about the side effect of such changes. Note: removing penalties is a global change. You cannot remove the Different Formatting Penalty  in some cases - i.e. the usage of superscript for linguistic reasons - but not in others. So if you remove the penalty, all tag discrepancies will be tolerated. Read more about this in this very interesting article or by clicking on the preview displayed below: