Meeting minutes
<gautierchomel> scirbe+
PR for providing placeholders for localization.
<gautierchomel> gpellegrino: it's a big PR but not so complicated. each placeholder is between brackets and there is a mention to remplace it by the corresponding <var>.
<gautierchomel> gpellegrino: matt itested to implement and found difficulties because updated translation are needed to make sure it works. Matt also revealed minor issues to fix with certifier informations and a question about date localisation.
<gautierchomel> gpellegrino: to sumarize, the proof of concept is ok, there are adjustment to do.
<gautierchomel> AvneeshSingh: I am as worried as Matt with the need to update parsers. That means there is no interoperability between the previous version and this one. Will we have to maintain too versions of the guide if implementation of loacalised plaecholders is not universal?
<gautierchomel> AvneeshSingh: i mean this has to be pushed asap so we can work on next changes.
<gautierchomel> CharlesL1: most changes concerns conformance and certification. I'll be happy to port to the EPUB techniques.
<gautierchomel> CharlesL1: what do we need to change in localisations? we should ask Rick the consequences for them.
Feedback - use of AI to create alternative descriptions - for a future update to guideline: Issue #458
<gautierchomel> Chris: should and how should we inform the use of AI in image descriptions?
<gautierchomel> gpellegrino: right now we recomand to inform that in the summary. It is not only for images alt text.
<gautierchomel> George: i agree the summary is the right place for now. We should also know if it is only automated additions or if there are curated prompts and human validation. we need to track if AI was used, how it was used and revised.
<gautierchomel> CharlesL1: easiest solution is to inform in summary but it would be very nice to have the info at the image description metadata level. Both for alt text and extended descriptions.
<gautierchomel> gautierchomel: the case also exist for infered metadata, if we automatically add it, how can this be identified?
<gautierchomel> Madeleine: we need to investigate that, there's already a proposal at ONIX regarding the AI voices used to generate audio version.
<gautierchomel> gluejar: on mass production at Gutenberg, we think it is premature to have a machine readable information for that, it could change in one or two years. Human produced alt text does not guarantee quality, in some cases AI would be better, specifically with formulaes. The summary is a good place to describe the practices for now. Let wait 5 years to see what a machine readable piece could be.
<gautierchomel> George: AI enters many areas in workflow production. From AI writen novel to AI qualirty control. It gonna be used everywhere. Probably in the summary we should describe how we use it.
<gluejar> would you put the same sort of information in the accessibility summary about the expertise of human created alt-text? Not a bad idea.
<gautierchomel> Chris: the case arised from publishers updating the backlist. They want to raise a caution flag like "this is not our usual quality". We think about naming the creator of the work of writing alt text, like a contributor.
<gautierchomel> AvneeshSingh: the audio piece is a very good use case because we ear the voice, is it a cloned one?
<gautierchomel> CharlesL1: when i published my book, amazon has questions on how much AI was used in the creative process (No, minor, heavy), and the system used. I don't know what they do with it but they collect that information.
<Chris> https://
<gautierchomel> Madeleine: I have started contacts to think about what and how schema should adress those questions. I will synthetise.
<gautierchomel> Chris: the link is a note form editeur about the use of AI, can be of use.
<gautierchomel> AvneeshSingh: let's continue the discussion in the issue tracker.
Any other business.
Next meeting?
<gautierchomel> AvneeshSingh: next call august 14th.