We would like to see a feature inspired by another platform: a QA criterion that detects unedited fuzzy matches. This is especially helpful for identifying cases where a small change to the source string results in a high fuzzy match from the Translation Memory, but the corresponding adjustment in the target has not been made. Implementing this would enhance quality assurance by ensuring all matches are properly reviewed and updated.
Hi @yeu_crowdin,
This feature already exists in Crowdin!
You can enable the Outdated translation (Translation predates source update and requires review) QA check in your Project Settings → QA Checks section. It’s at the bottom of the list. Just enable it and save the settings
Hi Roman,
Thank you for pointing this out. Unfortunately, I have not been able to get this criterion to generate QA warnings. Here’s an example of what I expect:
The original string reads “No limitation with that name was found on S2.” During the next update, the string is changed to “No limitation with that key was found on S2.” That turns the top TM suggestion from a 100% match into a 91% match. However, if the translation for the changed string is mistakenly kept, or 90+% matches are inserted by a translator, there is no pop-up when hitting the save button, reporting that an 7 month old translation is used.
Could you help me to get it to work? Thank you.
Hi @yeu_crowdin , so it’s expected since Outdated translation QA check flags translations that may need to be revised after their corresponding source strings have been modified
Hi Natalia,
Ok, I now managed to have this criterion generate a warning, but only by uploading an updated file and not keeping the approvals.
However, if a high fuzzy match is inserted and confirmed by a translator, no warning will ever be thrown because the criterion will always look at the time and date of insertion, but not at the TM match rate, right? In that case it will not act as an unedited fuzzy match check would in Trados Studio or memoQ.
Is there a feature in the API that would allow developing such a criterion, be it as a QA check template or an application? In its current form, this criterion would not have been able to flag an unedited high fuzzy match, so we need something that would.
Thank you for your understanding.
Hi @yeu_crowdin let us double-check this and come back
Hi @Natalia,
We’re wondering if there’s any update regarding this feature request / unedited fuzzy check mechanic?
Hi @yeu_crowdin ! Checking if there is any news with my team. We will get back to you soon!