Dear all,
Please find below today's meeting minutes.
Best,
Lucía
....
Attendance: Rodolfo, Mihai, Lucía. We have quorum
Administration
R: I move to the March 4, meeting minutes – https://groups.oasis-open.org/discussion/meeting-minutes-14
L: I second.
R: Meeting minutes approved.
Technical
XLIFF 2.2. Committee Specification 03.
https://groups.oasis-open.org/discussion/xliff-22-cs-files-uploaded
https://docs.oasis-open.org/xliff/xliff-core/v2.2/csd03/xliff-core-v2.2-csd03-part1.html
R: The ballot has passed. It has been approved as committee specification. I prepared the new files for our Oasis Administrator and I have contacted her a few days ago, so she can publish them. I have not heard from her yet (she was out of office until yesterday), Lucía, could you send her a reminder in a couple of days if we do not hear from her before?
L: Sure. I will do that.
(Rodolfo shows the new document (the date and the title has been modified). )
R: Once this will be published, we could announce that the new specification is out.
L: I think we need to update the main page of the TC before making the public announcement, Rodolfo could you do it?
R: Yes, I can. Once we have that, if we want to have it OASIS standard. If I remember correctly, we would need at least three statements of use from three different companies. I have to double check the rules with our OASIS administrator.
L: I think we should first wait for them to publish the committee specification and then move to the next steps.
XLIFF and IA.
L: I have been reading about the importance of XLIFF for training LLMs and I was wondering if you have any experience in this field.
R: It is very important for the LLM to have the bilingual format (XLIFF) for training process. That is why we were talking about creating a bilingual memory format based on XLIFF.
L: I did my PhD research on the importance of metadata in XLIFF for human translators, and I was wondering how important it is now for machines as well. I was thinking about inviting people that are working in this area so we can have a better idea of how XLIFF is being use in that scenario and what are their needs, we might be able to incorporate them if they are deemed relevant.
M: I can see how LLMs are getting better at translation.
R: I made some experiments myself. I decided to integrate ChatGTP in my tools but the first attempt did not provide good translation results. Then I took a different approach, by changing and specializing the prompt and then the results got better.
L: That is good to hear. So I will gather information on this topic and I will reach out to people who are publicly working on it by next meeting.
------------------------------
Lucía Morado Vázquez
Researcher and lecturer
University of Geneva
------------------------------