2025-09-24

Housekeeping

Welcoming new attendee, round of introductions

Action items from last meeting:

Follow-ups from last meeting:

• Primo Person entities at the view level

  • Melanie showed the UT Campus Collections view (which includes just cultural heritage collections from across campus) in UT Libraries Primo.  Not sure that this view is one where the Person card would be useful, but this illustrates distinct Primo views where it could be enabled.

• Bibframe records in Alma – Moving target.  No active Bibframe cataloging work happening at UT Libraries, so it’s challenging to prioritize exploring that.  Likely not going to be easy to really understand Bibframe in Alma until some of us are actively using that functionality.

  • Ex Libris is rapidly evolving what the support and infrastructure looks like for Bibframe, which is exciting.  But we’re still not quite there in terms of actively exploring it at our institutions.

• Action items following Josh’s presentation about Wikidata item change evaluations:

  • Sex/gender property being added by bots – possibility was raised in previous meeting discussion to submit a request that the sex or gender property in Wikidata be split apart.  Potentially something that this group could take action on.

    • What is the discussion about those properties and that bot happening on Wikidata?  How is that bot working?

    • Talk page for property P21: https://www.wikidata.org/wiki/Property_talk:P21 There have also been discussions about it in the LD4 Wikidata slack channel.

  • How do we handle cases where we disagree with edits like this that have been made to records we care about?  Would this be a situation where we would delete those statements, develop a bot of our own?
    
    

Announcements:

• We will need a new co-facilitator for this group, since Melanie is retiring from UT at end of year. Discuss at October meeting what group wants leadership, structure, and goals of future meetings to look like.

Updates from the field

ExLibris/Alma/Primo updates:

• Works search available in Alma, for institutions with BIBFRAME Support enabled.

  • Ties into questions from last meeting about how Bibframe is stored in Alma vs generated on the fly. More investigation tbd.

  • If new functionality is only available to institutions who have turned on Bibframe support, suggests this functionality may be shifting from experimental to more mature/robust.

• Current documentation about Linked Data and BIBFRAME in Alma/Primo:

  • https://knowledge.exlibrisgroup.com/Alma/Product_Materials/010Roadmap/Linked_Open_Data

  • https://knowledge.exlibrisgroup.com/Alma/Product_Documentation/010Alma_Online_Help_%28English%29/090Integrations_with_External_Systems/030Resource_Management/200Linked_Data

  • https://knowledge.exlibrisgroup.com/Alma/Product_Documentation/010Alma_Online_Help_(English)/Metadata_Management/005Introduction_to_Metadata_Management/BIBFRAME_Support

  • https://developers.exlibrisgroup.com/alma/integrations/linked_data/

  • https://developers.exlibrisgroup.com/alma/integrations/linked_data/BIBFRAME/

Recent UT iSchooler project at Library of Congress, involving LOC data and Wikidata items for historic landmarks/buildings and landscapes in testing of geocoder performance.

• Blog post includes a link to a Jupyter notebook with great details

• This project and methodology could be really useful for architectural research!  It’s currently difficult to get location data for resources in this collection.

• Also this project likely used some of the Wikidata items created or edited by the AAA!

Southern Architect project presentation

Devon, Hannah, Katie

postponed until October meeting

AI-assisted music cataloging project presentation

Hannah

Presentation:

• Project team: Hannah Moutran, Whit Williams and Corey Halaychik

• KUT Collection (part of Historical Music Recordings Collection), stored at CDL.  Collection is too large for catalogers to work through, or even identify the items that most need cataloging.

• Automated workflow:

  • Prep: scanning & labeling physical items

  • Metadata extraction & data normalization

  • Record matching using OCLC Worldcat Search

  • Analysis of results

  • Year and track verification

• Metadata generation: ask for structured data.  “Not visible” is given as an option – don’t invent things.  Some post-processing is done (e.g. numbers processed to exclude UT barcodes).

• That metadata is then used to create queries to the OCLC Worldcat API.  Use multiple different kinds of queries to get back the results that we want.

• LLM Assessment – Send the prompt, the parsed metadata, and the top 10 OCLC results, and get back what it thinks is the best match, a confidence score, reasons for the confidence, and other potential OCLC matches.

• Post-processing and validation – Check similarity of track list, year matching.

• Final artifacts and cataloging tools:  Spreadsheet and JSON file with all of the workflow information; guides for how to use these artifacts; sorting spreadsheet; low confidence tracking spreadsheet and other data; high confidence upload file for ingest into LMS

• Results assessment – grid of correct/incorrect (is it the right item?) and high confidence/low confidence (did the LLM correctly judge that the record was correct?)

• Next step: Alma integration – not there yet

• Review process is long and tedious – Developing a simple HTML page to make it easier to review, record decisions, and comments.

Discussion:

• One of the cool potentials of this project is the ability to identify unique or rare recordings in this collection (i.e. those that no one else has cataloged).

• Process of scanning the covers – had to make sure that the files were named with the barcodes, but it was challenging to keep it consistent.  One of the slowest parts of the process!

• Does the process use the AI-assisted metadata creation functionality in Alma?  No, this process has all taken place outside of Alma (including checking holdings in OCLC).  Next steps for the process will include integrating it with Alma.

Topics for ongoing discussion

• Entity reconciliation for place names and venues

• Partnering with UTL Research Data Services on Linked Open Data efforts

• Wikidata work - touched on in follow-ups from last meeting, see Housekeeping notes above.

• Unique Resource Identifiers for Archival Resources

• Hanlin’s Wikidata research Project

• BIBFRAME and Linked Data in Alma/Primo - touched on in follow-ups from last meeting, see Housekeeping notes above.

Wrap-up

• Next meeting will be October 22nd

• Action item review