Re: Zanata statistics and what it tells about us

On 7/13/19 8:40 PM, Jean-Baptiste wrote:
>
> Le 11/07/2019 à 07:10, Benson Muite a écrit :
>> Hi,
>>
>> In the unlikely event that the statistics are lower because of 
>> automation or because most of the translation has been done, perhaps 
>> this is a good thing!
> There is no automation in translation. Accessible translation memory 
> are not good enough, and the few translators with the skills don't 
> have the infrastructure to make any kind of automation.
This probably needs some co-ordination. Assuming this is something that 
many Linux distributions need, one might even try and see if this could 
be done under a larger body, such as Linux foundation.
>> It may be good to understand which applications are important for the 
>> typical user and need translation. 
> Important or not important application to translate varies really much 
> from one user to another. Selecting them may create useless conflicts ;)
If translation resources are limited, some coordination may be helpful, 
though of course individual motivation may differ and should be taken 
into account. Priority package groups are a good example.
>> Many of these applications are not translated on Fedora infrastructure.
>
> True, but if the priority package groups 
> (https://fedora.zanata.org/version-group/view/PriorityPackages) only 
> have 9 languages with more that 90%. There is still a lot to do.
>
>
Ok, this is useful.
>> As an example, it may be worth taking a closer look at Pontoon which 
>> is used for Firefox. The integration with machine translation can 
>> reduce the time needed for translation. As web browsers are important 
>> parts of daily computer use, this would likely have an important 
>> influence on overall user experience.
>>
>> Translation memory is helpful, but many platforms require this to be 
>> built up for each project. It would be helpful if there was a way to 
>> share translations between projects to allow for improved machine 
>> translation and translation memories - this still does require human 
>> input to produce a final high quality translation, but would reduce a 
>> lot of repetitive work. This might just be automated exchanging of po 
>> files, but maybe something more could be done. Software extensions 
>> that allow people to suggest improved translations, or indicate if a 
>> description is unclear would also be helpful.
>
> I understand, so if I rephrase:
>
> 1. As a translator, I want to be able to query a translation memory 
> containing a wide range of content so that knowledge beyond Fedora 
> community is accessible.
>
>   this means: include in our translation memory translation from 
> project out of the scope of the Fedora translation community.
Yes, many terms related to computers are also used for desktop clients, 
web browsers, computer games, word processing packages etc. all of which 
need translation.
>
> 2. As a translator, I get suggestion from an automation translation 
> engine based on a wide range of content to translate faster so that 
> translation time is reduced.
>
>   This means we have a tool we fed with our translation memory and 
> train it to have better result.
Yes.  In this case expect wider content related to computers, general 
content will typically give more suggestions that do not give correct 
context.
>
> 3. As a localization coordinator, I can automatically complete 
> existing translation using a translation memory using manual threshold 
> so that useless operation for translators are limited.
Yes.
>
> 4. As a localization coordinator, I can automatically complete 
> existing translation using an automation translation engine manual 
> threshold so that useless operation for translators are limited.
Yes.
>
> 5. As a translator, I can ask for an automated translation of a file 
> and review suggestion so that starting a new translation is faster.
Yes - may mean can get a wider community doing translation review.
>
> 6. As a localization coordinator, I have a tool to handle non-fedora 
> translation content so that translation memory is continuously update.
Yes.
>
>
> Does it match what you wanted to say?
Yes. Note that Weblate does have support for integrating machine 
translation engines (https://docs.weblate.org/en/latest/admin/machine.html)
>
>
> Jean-Baptiste
> _______________________________________________
> trans mailing list -- [email protected]
> To unsubscribe send an email to [email protected]
> Fedora Code of Conduct: 
> https://docs.fedoraproject.org/en-US/project/code-of-conduct/
> List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
> List Archives: 
> https://lists.fedoraproject.org/archives/list/[email protected]
_______________________________________________
trans mailing list -- [email protected]
To unsubscribe send an email to [email protected]
Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives: https://lists.fedoraproject.org/archives/list/[email protected]