Start a new topic

QA intra-segment consistency?

Is there already something like intra-segment consistency (internal consistency) possible?


E.g.:


Segment 1: AAA = TAAA

Segment 2: AAA BBB = TAAABBB

Segment 3: AAACCC = TAAA CCC

Segment 4: AAA xyz AAA BBB abc AAACCTAAA hij TAAABBB klm TAAA CCC


In plain language:

For source and target segments pairs where one side only contains one word (no spaces), make sure that the translation entered here, is being used in all other segments.




Actually, inter-segment consistency, I guess?

It is still morning and I did not have my 2nd coffee yet, but isn't this simply to handle with TB or fragments consistency?


Just asking …


Igor, Why aren’t the errors in the fourth segment not reported? / A possible approach: filter on segments that don’t contain any spaces in source and/or target. Import the filtered segments into a memory. Either use this for QA Fragments (but it’ll probably fail, just like above). Or convert the memory to a glossary and use this for QA Glossary. / Anyone knowing a shorter route?

In times of heavy MT-assisted translating, this feature may well evolve to a crucial quality factor, since MT systems like DeepL until now have no way to guarantee inter-segment terminological consistency.


This is where we professional translators come in :).


So, an integration with the MT engines would be nice. Kind of advanced teaming of auto-assembling with MT.

Your AAA, BBB, CCC examples are a bit unclear to me but you should have the corresponding fragments as fragment matches in your translation memory to be able to perform the consistency check.

>Your AAA, BBB, CCC examples are a bit unclear to me but you should have the corresponding fragments as fragment matches in your translation memory to be able to perform the consistency check.


Yeah, I was a little afraid of that. Tried to make a generic example. Will replicate with real-life words.

So, you are saying that it's not enough to have the segments 1-3 as segments in the TM to make the QA fragments work? I should also add them as fragments, manually (I always thought that CafeTran Espresso 2019 considers everything that has been saved as a segment in the TM, also to be a fragment ...).

It might work for segments too but you seem to be nesting segments within fragments (or the other way around). It may not work for fragments nested in, or being a part of, other fragments.

Confirming this QA action works as expected when the TM used for this check is set as a fragments-only memory (termbase) and not as a segments memory as well.


Good to know!


2 people like this

Me too. Now I understand.

New test of QA Fragments consistency. It is important that:

  • No sentences or longer segments that span several fragments are imported in the TM for Fragments consistency.
  • The TM should not be set to Segments, only to Fragments (and Fragments consistency of course).
This is the project and TM Fragments:

image


I get these results for QA Fragments consistency:

image


And:

image


Question for Igor: Why aren't the fragments 'linke Seite' (= linkerkant) and 'obese Seite' (= bovenkant) not reported? Is it because they consist of two words at the source side?

I converted the TM for Fragments to a glossary for QA and duh, the same result: linke Seite and obere Seite are not reported as inconsistent.


image


Please check your option in Edit > Prefrences > QA tab > Terms consistency check: Prefix/Exact matching.


1 person likes this

Greatzzz


image


> The TM should not be set to Segments, only to Fragments (and Fragments consistency of course).


As you start the "Terms consistency check" action, the pop-up dialog appears with the TMs listed, regardless of their type (for Segments or Fragments).  Just tick/untick the TMs for this QA check. In certain system look and feels (e.g GTK 3 in Linux), the unchecked/unticked state is blank.

Login to post a comment