Following on from my earlier post on creating a new translation memory (TM) in SDL Trados Studio, here are some tips on using penalties and filters. As your TM grows bigger, some results may be less useful than others; likewise, if you use several TMs in a single project, some may be less relevant than others. Read on to learn how to make sure you see the best matches first in your TM look-ups.

Penalties

Apply a penalty to an entire TM

If you have several TMs assigned to a project and want to give lower priority to matches returned from one of them, assign it a penalty in Project Settings (F2) > Language Pairs > All Language Pairs > Translation Memory and Automated Translation. In the screenshot below, I’m translating a file for Client A, but have translated a similar text for Client B. However, I want to make sure that Client A’s matches are returned first, so I assign Client B a 2% penalty:

(Click any screenshot to open it enlarged in a separate window)

A 100% match in the penalised TM will now be shown as 98% in the Translation Results window, which will alert me to check it carefully to see if I can use it as is for my current project.

Penalties also affect fuzzy matches (so a 90% match would be shown as an 88% match in the above example), but that’s fairly irrelevant. Their value really lies in stopping a potentially unreliable 100% match from being pretranslated at the file preparation stage.

Apply a penalty to specific segments in a TM

By default, Studio applies a 1% penalty when it encounters formatting differences, multiple 100% matches and results from alignments. You can change these penalties for all future projects in General Options by going to File > Options > Language Pairs > Translation Memory and Automated Translation > Penalties, or change them in a current project by following a similar path in Project Settings (F2).

You’ll already be familiar with these default penalties in practice when, for example, the only difference in a segment is a pair of formatting tags:

Note that Studio doesn’t impose a default penalty to auto-localization or text replacement in the above options. I’ve tried applying an auto-localization penalty to overcome Studio’s problem with long date formats (it doesn’t handle days of the week correctly), but it’s not ideal because the penalty is applied to numbers as well.

The text replacement penalty, however, is useful. This affects variables, acronyms and alphanumeric strings. I apply a text replacement penalty to compensate for Studio misclassifying alphanumeric strings. Otherwise, for example, Studio would simply transfer ALFA-1 to my target segment as a 100% match. (In English, it needs to be translated to ALPHA-1.)

Filters

Filters are for applying additional penalties to specific conditions in a TM that are defined by customised field values and system values.

I have a regulatory TM with official translations from EU legislation and similar sources. I added an “unofficial” translation of a Spanish decree, published on the Spanish Ministry website, and identified it as such in a text field. I apply a penalty to that resource to warn me that it may be unreliable.

Take care with the syntax when defining the penalty, because it applies to non-matching translation units. Therefore, the NOT tick box must be marked to apply a penalty in the case above. Note that you can combine conditions with and and or operators, and add more than one filter penalty to your TM.

Penalties applied to system values may be useful, for example, to mark down segments created by another translator or those translated more than five years ago:

Hard Filter

The Hard Filter option is similar to applying a 100% penalty. Any segments that meet these criteria will simply be excluded from search results.

Conclusions

Applying a penalty to one TM in a multi-TM project is a valuable feature, which I use heavily in my work. Penalties and filtering of specific values and fields is more complex, but worth understanding and practising so you’re ready to use these features when the need arises.