Submitter's message ActionItems; none
===============================================
Minutes of the OASIS DITA TC
Tuesday, 17 March 2026
Recorded by Nancy Harrison
link to agenda for this meeting:
https://github.com/oasis-tcs/dita/wiki/Previous-agendas
Attendance:
===========
Robert Anderson, Stan Doherty, Kristen Eberlein, Nancy Harrison, Scott Hudson, Bob Johnson, Eliot Kimber, Zoe Lawson, Christina Rothwell, Eric Sirois, Leroy Steinbacher, Dawn Stevens
Business
========
Regrets:
1. Minutes
3 March 2026 (Harrison, 04 March 2026)
https://groups.oasis-open.org/discussion/dita-tc-minutes-3-march-2026-uploaded
- Kris moved, 2nd Scott, approved by TC
2. Announcements
- Dawn; at ConVex, we'll be celebrating 20 years of DITA, so we'll be having 'DITA Jeopardy' questions. If you have a list of 5 questions that would go in a category. e.g. about maps, elements, TC, etc., please send them to me. Conference is april 10-13.
- Bob; Boston DUG, April 8 will have Stan talking about using AI with DITA
- Stan; AI can give reasonably good help for learning some things about certain aspects of DITA, quesstion is 'where is it not helpful?'
- Kris; before call started , Eliot was giving an update on what he's doing on AI, so maybe item #4, talking more in detail with Dawn about questions.
3. Committee notes effort (03-17 updates)
- Leroy: DITA 2.0 Migration Guide - in progress
- Stan: Introduction to DITA for Writers - in progress
- Stan; making progress on 'Intro to DITA', expect to make more progress before Convex, so we'll have a little handout to recruit for TC, see if we can come up with a flyer
- Leroy; I've downloaded 'Migrating to DITA' from github, will go thru it, make notes. couple of keyrefs not resolving, we'll try to figure it out.
- Stan; I'm working in 2.0, and it seems to be working really well. except for glossref.
- Leroy; same from me.
- Kris; i'm glad AEM Guides will be supporting newest Java, which is required for DITA 2.0. until recently, it ddin't, and you could only use 1.3.
- Committee Notes.
https://github.com/oasis-tcs/dita/wiki/DITA-TC-Committee-Notes
- Updated description of mission
https://groups.oasis-open.org/discussion/updated-mission-description
4. Eliot; update on AI and dita
- Eliot; I'm setting up connectors so users can use Claude with Positron. I had Claude write the file, then convert it to DITA., asked it where it should go, it updated the map and created the topic, using a Markdown file as a guide, It then created the topics and created PNG placeholders for the images with descriptions, then created keydefs for each image and put them in the map. Really good stuff... It didn't get the image refs first time, because it didn't know it needed to use keyscopes, but it saw the Schematron error messages and fixed its own error. So now I just have to create the screenshots, then validated everything. we're stting up NTP connections to internal repos and local file systems. The AI seems to learn our markup rules by just looking at the content. ServiceNow has its own internal LLMs now, so nothing goes outside our firewall.
- Kris; since you're working on Github, and Positron has access to your local repos, it's doing it in your local copy, which you then need to commit?
- Eliot; right, but Claude can also do your Github commands for you. I used it to analyze Git history to find the last author who modified a topic; it was able to figure that out in a reasonable amount of time. so I could do authorship assignment we need. Then it wrote a doc that explains our procedures; I no longer do any programming.
- Zoe; with AI help, I got a reasonable conversion of an 900 page Framemaker book to DITA, with just some internal xrefs problems.
- Kris; There are places where AI doesn't work very well, w.g., we have enterprise-wide Copilot, you can add ChatGPT, also most of Mayo's other content. When I gave it a prompt to get best practices for page breaks in tech docs, I got back a wonderful looking doc, but when I checked the sources it recommended, it fell apart, because all the sources were to very derivative sources, like paragraphs abstracting content from Chicago Manual of Style. So, it's hard to use for big picture, synthetic questions,, and also, a lot of the sources you would need to use aren't publicly available on the web.
- Bob; that ties in to something I've been wondering about, working with equipment mfrs. If an org is developing stuff for cad-cam systems. I'm not sure AI can do that very well. AI is great for software, but once you go to hardware, not sure how useful it is. Also, there's a question about governance; if you're working in an internal LLM, when external content is being updated, how do you update the LLM that was trained on it?
- Eliot; that's part of NCP approach, instead of training AI on your content, you retrain it on new content. NTP lets you say 'use this dataset' as though you had.
- Kris; if your content is locked into CAD, how do you use AI?
- Eliot; AI systems can do some interpretation of images, but can Claude read a file and tell you anything useful? I'm not sure
- Kris; I've been impressed with how much AI can do with images, deconstructing layers and reading included text.
- Scott; we've been impressed as well. pretty good at generating alt-text.
5. planning for Convex jeopardy game.
- Dawn; any ideas?
- Bob; what about accessibility question.
- Kris; what about what are problems that DITA was designed to be an answer for?
- Dawn; cost justification?
- Kris; specialization? doc-type shells? I've been working with specializationss of 'topic', not concept/task/ref, but the task topic is a really thing of beauty....
[folks added their jeopardy questions to the chat for the meeting]
close 11:53 am Et
ActionItems: none
-- Ms. Nancy Harrison
---------------------------------
Nancy Harrison
Principal, Infobridge Solutions
Nancy Harrison (Personal)
---------------------------------