barefootliam

I’ve been trying to post (not always in the same place) about the usefulness of text #markup for documents - #XML especially - using the tag #MarkupMonday each time.

The goal is to spread awareness that XML, and declarative markup, labelling of parts, is in active use, and yes is worth doing, in the right place.

There was some push-back on the xml-dev (reborn) mailing list against XML being useful in a world with LLMs in it, which was discouraging.

But i stick by it: XML is not dead! 😉

May 20 at 2:49amWeb
Show threadbarefootliamMay 20
(if you are wondering, Tuesday is the new Monday, it was too hot here yesterday! Maybe it’s #TagsetTuesday today!)
Show threadDrScripttMay 20

@barefootliam why would XML not be useful for LLMs?

I’d think that semantic markup would be a good thing for LLMs.

IMHO XML is more useful than JSON.

Show threadbarefootliamMay 20
@drscriptt that actually seems to be what a lot of people have found, yes. Many LLM scrapers remove markup before ingesting the text, but XML does seem good for prompts.