How to use TM to perform QA check?

General Discussion
1

Hi there,

I think it would be really helpful to use TM to perform QA check to check 90~95% match rate sentences.
Is there any way to do that?

2
  1. You should export the initial TM to tmx (i.e. before translating or reviewing).
  2. Add the .tmx file as Ongoing Translation to your Xbench project.
  3. Add the bilingual translated/reviewed files as Ongoing Translation to the project.
  4. Go to QA tab > Basic check group.
  5. Enable the Inconsistency in Source checkbox.
  6. Click Check Ongoing Translation.

Those fuzzy segments that been confirmed but not modified will be reported as inconsistencies in source when checking the TM and bilingual files as Ongoing Translation.

1 Like
3

Hi omartin,

Thank you for your message.
ā€œInconsistency in Sourceā€ would only find those have same target sentence.
What if I would like to check if certain target sentences follow the TM consistency?

Example:

[TM]
Source: Jā€™ai vu un stylo sur la table.
Target: I saw a pen on the table.

[Translated File]
Souce: Jā€™ai vu un pomme sur la table.
Target: I saw there is an apple on the table.

From above, you may see the target sentence does not follow ā€œI saw XX on the tableā€.
Is there any way to check the situation?

4

To get inconsistencies in source, the target text must match.

To detect segments that may have only a different word but have not been translated in the same style, I would use a checklist such as the following one:

Source: "Jā€™ai vu un <[:letter:]+> sur la table\."
Target: -"I saw an? <[:letter:]+> on the table\."

5

Hi,
Thank you for your regex!
Itā€™s very helpful. However, I need to adjust the pattern based on specific sentences.
I look forward to having the similar feature like I proposed in the upcoming updates. :slight_smile:

6

Is it possible to set xBench so that it detects inconsistency between segments even if their match is lower than 100%?
Let me provide an example.
I would like. e.g., to have xBench telling me that there is an inconsistency between:

  • ā€œI eat a red appleā€, and
  • ā€œI eat that red appleā€.
7

Hi Davide,

To get inconsistencies in target, the source text of both segments should be the same.

To get inconsistencies in source, the target text of both segments should be the same.

If the target and source text of both segments is different, this issues can only be checked via checklist entries.

8

Thank you! Your advise helped me :wink:

9

Thank you very much for your feedback.
Davide