bugGNU gettext - Bugs: bug #55871, Integrate with machine translation...

 
 

You are not allowed to post comments on this tracker with your current authentication level.

bug #55871: Integrate with machine translation tools

Submitter:  Bruno Haible <haible>
Submitted:  Sat 09 Mar 2019 04:57:08 PM UTC
   
 
Category:  Translator tools Severity:  1 - Wish
Item Group:  None Status:  None
Privacy:  Public Assigned to:  None
Open/Closed:  Open

Jump to the original submission

Sat 18 May 2019 10:16:11 AM UTC, comment #6: 

I agree that gettext could provide a way to pre-populate the translation. It could be from a MT provider (by providing the info required by the service API) or from a local TM file in a supported bilingual format (PO/TMX ?). That would be tremendously helpful.

Jean-Christophe Helary <suzume>
Sat 18 May 2019 05:57:43 AM UTC, comment #5: 


> At least some of these services, under GPL,


The GPL is a license for software, it isn't applicable to services.

> can be used as tools, without some of the concerns of network services.


Whenever someone uses other people's servers to do their own computing, some concerns remain: for example, the programs on the server can't be modified by their effective users.

Ineiev <ineiev>
Sun 10 Mar 2019 12:00:16 AM UTC, comment #4: 


> I recommend using Pology (the ‘pomtrans’ binary)

Thanks for mentioning this! I have now added a pointer to Pology to the gettext manual.

> I have successfully used Pology/pomtrans (with the offline Apertium backend) for initial translations of PO files.

As far as I can see from http://wiki.apertium.org/wiki/Main_Page, Apertium is good for languages with relatively few speakers, for example to create an Occitan translation from a French translation. Unfortunately, for a translation from English currently it offers only the target languages Spanish, Galician, Catalan, and Esperanto. This means, there still is a long way to go.

Bruno Haible <haible>
Group administrator
Sat 09 Mar 2019 10:38:36 PM UTC, comment #3: 

I don’t see a need to implement this in gettext. There already exist tools that can translate PO files using machine translations tools/services. I recommend using Pology (the ‘pomtrans’ binary), which you can read more about http://pology.nedohodnik.net//doc/user/en_US/ch-misctools.html#sec-mimtrans

I have successfully used Pology/pomtrans (with the offline Apertium backend) for initial translations of PO files. The resulting files need careful proofreading, but having a rough machine translation available saves time.

Karl Ove Hufthammer <huftis>
Sat 09 Mar 2019 06:09:03 PM UTC, comment #2: 


> services raise different concerns

Yes, services raise issues such as:
- Who is in control of the neural network / data?
- Is it biased (implicitly through the input it was fed with)?
- Does copyright apply?

At least some of these services, under GPL, can be used as tools, without some of the concerns of network services.

Bruno Haible <haible>
Group administrator
Sat 09 Mar 2019 05:16:06 PM UTC, comment #1: 

Generally speaking, services raise different concerns,
https://www.gnu.org/philosophy/network-services-arent-free-or-nonfree.html


You may want to discuss this with rms.

Ineiev <ineiev>
Sat 09 Mar 2019 04:57:08 PM UTC, original submission:  

Machine translation is getting better over the years, thanks to advances in AI: machine translation based on neural networks and deep learning is significantly better than machine translation based on statistical models.

An approach that works in practice is machine translation followed by post-editing https://en.wikipedia.org/wiki/Postediting .

The GNU gettext tools, in particular msginit, ought to have an option that allows to easily use free machine translation tools.

We should only integrate to free (GPLed/LGPLed) machine translation software, because the GNU maintainers guide https://www.gnu.org/prep/maintain/html_node/Ethical-and-Philosophical-Consideration.html says "A GNU package should not recommend use of any non-free program".

Fortunately such free machine translation software exists nowadays, see https://en.wikipedia.org/wiki/Comparison_of_machine_translation_applications . For another list, see http://docs.weblate.org/en/latest/admin/machine.html . There is also ModernMT https://github.com/modernmt/MMT .

Bruno Haible <haible>
Group administrator

 

(Note: upload size limit is set to 16384 kB, after insertion of the required escape characters.)

No files currently attached

 

Depends on the following items: None found

Items that depend on this one: None found

 

Carbon-Copy List
  • -email is unavailable- added by suzume (Posted a comment)
  • -email is unavailable- added by huftis (Posted a comment)
  • -email is unavailable- added by ineiev (Posted a comment)
  • -email is unavailable- added by haible (Submitted the item)
  •  

    There are 0 votes so far. Votes easily highlight which items people would like to see resolved in priority, independently of the priority of the item set by tracker managers.

     

    No changes have been made to this item

    Back to the top

    Powered by Savane 3.13-f8d8.
    Corresponding source code