Metadata Call 2026-02-24
Time: 2:00pm-3:00pm Eastern / 1-2 pm Central / 11 am-12 pm Pacific
Community Notes: https://docs.google.com/document/d/1G8wmi9R3q1uHatQLa3JfmlVr1-laTO5vgYzUZz-B7QU/edit?tab=t.0
Moderator(s): Annamarie Klose
Notetaker: Emma Beck
Attendees:
@Annamarie Klose, Ohio State University
@Emma Beck- University of Louisville
@Benjamin Riesenberg - U. of Oregon
@Anna Goslen, UNC-Chapel Hill
@Tiffany Chan, University of Victoria
@Morgan McKeehan(she/her/hers), Oregon State University Libraries & Press
@Juliet Hardesty, Indiana University
@Sudha Anand, Indiana University
@Nick Steinwachs, Notch 8
@Sarah Proctor, Notch 8
Steve McDonald, Tufts
Agenda:
Samvera updates
Bulkrax updates are in progress - last demo is tomorrow with Notch8. Test it at (demo.hykudemo.org/) contact Nic Don for access
Accessibility presentations - Video recording and WebVTT captions file
Patrick Burden, West Virginia University - Postponed due to a scheduling conflict.
Stacy Snyder, University at Buffalo - recorded
Accessibility Coordinator for the Libraries at the University of Buffalo
Content in Digital collections not accessible
Generative AI to create alt-text for images
Tested 3 gen ai services
4k(?) images from various categories of content. Compared the various alt-text and description and ChatGBT was the best
Using gen ai for OCR and text
Seeing lots of improvements since originally starting this work
3 students doing DC remediation
Steve McDonald, Tufts University
Library and Library IT group has been doing sprints to take on projects with AI. Running in parallel with one happening across the university
Experimented with transcripts creating video and audio using whisper
Library volunteers are doing validation of the work
Working on 30 videos ranging 3 minutes to a couple hours
OCR is the new gen ai sprint - Institutional Books Initiative
Using Tesseract for OCR
Limited test produced few hallucinations
Annamarie Klose, Ohio State University
44k records in digital collections
Focusing on publicly available content, then institutional records, then private content
File level accessibility support
Supporting accessibility through: ai, repurposing existing summary and description values, outsource to venders
Have to use University approved venders
Using Human in the Loop review
Claude sonnet workflow
Adding to alt-test “detailed description can be found in the description field”
Tracking progress in airtable
Relying on those who submit content to update their content accessibility
Questions:
Did I just hear that OSU’s accessibility features are going back to Hyrax main?
YES
Nothing up in github yet. Have shared information with the Hyrax interest group
What type of documents are people using for their OCR remediation and testing?
UB - worked on handwriting from 21st century, and printed items from 1800s forward. Better results from ChatGBT than out of specific OCR software. PDFs are different can of worms. Uses ABBY Fine Reader for PDFs.
Tuffs - focusing on doing OCR for books that are digitized, almost exclusively out of copyright material
OCR with handwritten text - some recommend ChatGB, but not great results all around
Various tools on how to convert to PDF on demand. One idea:
Presentation from Patrick at NISO last week. Shared notes
Various policies vs risk for accessible content? Archives exception
IU: We have an on-demand accessibility remediation form, and most of our requests have been for time-based content. For PDF accessibility the remediation group was trained on Equidox which has a limited number of licenses share across IU campuses
UB - digital collections and ir are living objects, so don't consider them archives. University archives use presevica and using the exception there. Some overlap between DC/IR and preservica
Tuffs - all digital materials, acquired, adopted, developed, or updated after April 19th must be accessible. Anything before then must be made accessible upon request. Collection uploaded before that April date does not need to have to be accessible
Where does information literacy come into play?
Humans do better at editing than starting from scratch. AI content to create a chart or other content than starting from scratch
Open discussion (time permitting)
Next meeting: March 24, 2026