Change "." (full stop) to "।" (dandi) in Content Translation when people nare translating from any language to Punjabi.
Description
Description
Related Objects
Related Objects
Event Timeline
Comment Actions
This seems to be an issue with the external machine translation services. I reported this to Google Translate team, feel free to add more details there or subscribe to the report to get further updates.
We try to avoid making language-specific fixes on top of the output of Machine Translation Services since these get updated over time and patching on top of them tends to be unreliable. Instead, we plan to have a more general system to learn from user corrections (T96165), saving users from correcting the same problem again and again.
Meanwhile, there are a couple of alternatives for editors to correct these issues. For example:
- Individual editors can use the search and replace functionality (ctrl+f) to replace recurrent problems in the translation.
- Communities can create bots to fix common issues in translations after they are published. Those bots can target articles tagged with the "contenttranslation" edit tag or be applied more generally.
Comment Actions
@satdeep_gill although the status to the upstream bug shows as "assigned", the issue seems not to occur in some examples I tried with Google Translate. Can you confirm whether it i working from your side?
In addition, Punjabi is also supported by MinT now. Can you confirm whether this issue is present for the translations with MinT (as well as any other input about their quality)?
Thanks!
Comment Actions
MinT(en->pa served by indictrans2 model) does use Devanagari Danda.
There was an extra space before Danda which I fixed with https://gerrit.wikimedia.org/r/c/mediawiki/services/machinetranslation/+/935538
Comment Actions
This seems to be solved now. Thank you for your support. I think that the Mint translation is better. I will get with the community to test in an upcoming in-person event.