Alternative vetting model: agree and sign off rather than review and proofread

The current review system is kind of anti-social: one person sais he has reviewed the string (meaning no obvious mistakes were made in spelling grammer or word order) and that one person locks the string from input from all others. Yes they can still add suggestions but I think we all know in the back of our mind that suggestions don’t get the same amount of attention as existing translations.

This somewhat kills quality improvements and willingness from new translators to take part in projects.

I think a different vetting model would work better for some projects:
Approve and sign off

Instead of a review button which locks a string, projects which use the approve and consensus model will have an “approve” button which does not lock the string. The string has a counter of how many reviewers have approved the string, and there is a list of all the members who approved this translation.
Project maintainers can set a minimum number of votes required on strings on the project before the string the translations get used.

Only language coordinators can mark a string as “final approval”. This locks the string from anyone but language coordinators for that language.

This is not a voting system where the majority rules. The language coordinator(s) rule(s). He or she guarantees quality and consistency in the translations, reviewers back up the decision making.

Project maintainers can set a minimum required number of approvals before a coordinator can press the sign off button, this guarantees that reviewers get to review first.

Language coordinators can still add new members to a language team like is already the case.

What this model would do better:

  • promote discussion and comparison between suggestions to reach the best translation
  • value new input, don’t reject new input as mearly suggestions
  • one or a few people sign off on every string which improves consistency
  • the language coordinator leads the language team and is the spokesperson for the project maintainer
  • strings get multiple reviews by different reviewers, rather that one review by whoever comes along first.

This new model could coexist alongside the existing review and proofread, a project maintainer could choose which vetting model they want to use on a per project basis.

2 Likes

Hi Meteor0id,

Thank you for using Transifex and for submitting your feedback!

We really appreciate your input, since this is the best way to help us improve Transifex and we hope that you will keep on enjoying the whole translating experience.

I have already forwarded everything you pointed out to the relevant team and they will review your suggestions, but please keep in mind that implementing new features or applying any changes might take some time.

Please, feel free to contact us with any further feedback or questions you might have!

Kind regards,
Cesar

Guess I was hoping for a bit of dialog between Traniffex designers and translators, rather than an internal forward and waiting for what might or might not happen 5 years from now.

1 Like

Hello Meteor0id,

I apologize if I gave you the wrong impression, Our team keeps working hard to improve our tools and provide a better translation experience to our users :slight_smile:.

In the meantime please check our documentation on crowdsourcing project type which is available on Premium and up plans and let me know if this is what you are referring to.

  • This type of project relies 100% on translation votes which means that the translation with the most votes fulfills the minimum number of votes required is automatically set as the final translation string. As a result, language coordinators do not have the option to choose another translation instead.

We could evaluate and consider your suggestions for further improvements

Also If you are interested in helping us shape product roadmap, You can always request to join our Transifex user panel - User Panel - Transifex

Let me know if the above information helps

Yes please, because the model I am suggesting is different from the current models.

A voting system aka crowdsourcing system has the downside that it doesn’t lead to much optimization or enhancement. For apps with a fast development cycle (I am thinking Signal messenger for instance) the translations are continuously being updated and rewritten to perfection. In a voting system that kind of perfection would not be possible, as new suggestions would have to compete with older suggestions which may have already received many votes. On top of that getting votes would require constant lobbying and discussion between translators and reaching consensus on even the smallest change, which is too much discussion and no one would attempt it.

What I am suggesting is a new workflow option of which the key difference is that each language team has one person who takes charge and responsibility of the quality, and everyone can at any time suggest changes.

The project maintainer picks one or a few translator(s) per team as the language coordinator and that coordinator gets the final say on each translation, but is constantly challenged by the other translators to approve better a better translation on the same string.
The idea is keeping each other sharp, keep improving which words you choose and which word order you’re using.

It’s a complex idea but if the Transifex people have any interest in this maybe we should have a call about it some time.