Metadata Call 2026-07-28

Samvera Community Wiki


Metadata Call 2026-07-28

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

 

2026-07-28

Moderator(s): Annamarie Klose

Notetaker: Emma Beck

Attendees: 

  • Jessica McMillen - WVU

  • Sarah Almond - WVU

  • LaRita Robinson - Notch8

  • Heather Greer Klein, Samvera

  • Emma Beck - University of Louisville - she/her

  • Rachel Howard - University of Louisville

  • Benjamin Riesenberg (they/them) - U. of Oregon

  • Morgan McKeehan (she/her/hers), Oregon State University Libraries & Press

  • Steve McDonald - Tufts University

  • Anna Goslen, UNC-Chapel Hill

  • Sarah Cogley - University at Buffalo

  • Christine Peterson, CPC

  • Rachel (she/her), Georgetown University

  • Julie Hardesty, Indiana University

  • Nic Don Stanton-Roark (he/him), Samvera/PALs

  • Sudha Anand, Indiana University

Agenda: 

  • Samvera Updates

  • SMIG meeting time

  • Nick (LaRita) @ Notch8 project feedback

    • Compound metadata - can pick multiple options for a field

    • Guided deposit UI - controls everything that was previously in the deposit, much of the same functionality just different look/feel. Adds access to set file level metadata and permissions. Currently in the Hyku staging and testing sites

      • Demo at Hyku Interest group on guided deposit on 7/30 11:30-12:30 EST

  • Alt-Text AI presentation from Olivia Wood, Ohio State University

    • Recording, WebVTT file, and Slides

    • Claude sonnet is effective for photographs 

    • Criteria for a good alt-text

    • Some collections were too extensive to be used

    • Slides - https://kb.osu.edu/handle/1811/107417  

    • How is the alt text provided in the interface? Is it part of the UV or on the thumbnail list of images or both or somewhere else?

      • In two places. In items list and somewhere in the UV

      • Some customized UV work

    • Have you experimented with having AI assess AI description and flagging for expedited human review?

      • Have not - in a related research project AI confidence levels can be deceptive 

    • And what kind of token budget do you have for Claude? Does your library get budgeted for that?

      • 1600 token each

    • Oregon Digital has also piped alt text into the UV right-hand nav and into WORKS AND FILES -- I don't have any of these development details of course

    • Here’s a hyrax ticket to add any info re: alt text in UV https://github.com/samvera/hyrax/issues/7377  

    • Ohio State University’s UV support for Alt-Text: Chris Sellers: “The source code will never show you the Alt Text, because the page’s source code represents only what’s present in the HTML markup at initial render time, i.e., in response to the browser’s initial call to https://library.osu.edu/dc/concern/generic_works/j6731j34d . The IIIF viewer is populated by the manifest, at https://library.osu.edu/dc/concern/generic_works/j6731j34d/manifest, in one of several post-page load Ajax calls that the IIIF viewer kicks off. When I loaded the page just now, the Network panel (in Chrome’s web inspector) shows the source code downloading to the browser at 1.16ms. It’s the first network call on the page, and it’s what the browser shows when you select “View Page Source.” The manifest is about the 60th call, and its download at 1.85s, an eternity later. With it in hand, one of the viewer’s JS files—presumably uv.js—uses the manifest’s data to hydrate the viewer. To see what the user sees (and hears), look in the Web Inspector. In Firefox, select the “Inspector” tab. If in Chrome, Safari, or Edge, select the “Elements” tab. You’ll find the Alt Text in our iframe, in .uv-seadragon-extension > .mainPanel > .centerPanel > .content > .viewer (where the dot prefix means a <div> with that class).”

  • Open discussion:

    • language tags and language support - Next month

      • Alignment issues for Arabic and non roman languages 

      • Georgetown will reach out to their developer to see if we can get a demo at some point 

      • n8 - had a client wanting titles in multiple languages and we were going to store it in a compound metadata field. Reading the metadata and mapping to the selected language would have taken more work.

      •  

    • Lots of request to hear more about compound metadata - especially how it will be stored in the back end

    • separate future topic, I think the functionality of Flexible Metadata/M3 has continued to expand and would be worth an updated demo