Updates from the field | | ALA Linked Data session highlight (paloma) Attention shift to AI – focus of more technical presentations has moved from LD to AI OCLC sneak peek of a BIBFRAME editor OCLC collaboration with West Virginia University Libraries to pilot a NACO semi-automated workflow through the use of Meridian, WorldCat Entities, and a WorldCat ontology 2 MARC format for Authority Data mapping (The Mothman properties) Equinox collaboration with the King County Library System (WA) to research implementation of BIBFRAME in public libraries, and how Evergreen can make use of metadata that originates form BIBFRAME
ExLibris/Primo update LD4 Conference takeaways Session recordings are available in a YouTube playlist Josh shared notes on the ld-learning list about the Performing Arts session The LD4 Art & Design Affinity Group is proposing to develop a separate affinity group for performance- and time-based records, such as theatre, dance, music, film, etc. https://ld4.github.io/art-design/ NYPL (Emma Clarkson, Samie Konet) , have been developing data models that distinguish and link theatre companies, performances, theatre buildings, and performance halls within buildings - https://www.wikidata.org/wiki/Wikidata:WikiProject_New_York_Public_Library/Projects (see the "Enhancing Off-Broadway theatres" project in particular) MLA LDWG ("Ludwig") group - works on data modeling for music-related records https://cmc.wp.musiclibraryassoc.org/mla-linked-data-working-group/ Harvard Film Archive Project - data modeling for film/video works including recordings of performances and events, and the film reel or video cassette object and its variants, https://www.wikidata.org/wiki/Wikidata:WikiProject_Linked_Data_for_Production/Harvard_Film_Archive_Project Our discussion
Discovery Affinity Group session: opportunities and questions about discovery and linked data slides and discussion notes Discovery Affinity Group charge and running meeting notes Linked Data in Systems for Discovery spreadsheet Discusion around possible directions discovery take with linked data, projects or products that represent good examples of using linked data for discovery, and which important projects or areas of exploration would attendees like the Discovery Affinity Group to cover. Our discussion Hannah: Using AI to support linked data discovery – AI can help with creating SPARQL queries. Also, AI extraction of entities has been really successful, lots of possibilities for that. Converting triples into vector embeddings to support AI agents (using linked data to feed into LLMs) – but questions about whether there is a high enough volume of triples to make that useful, and whether the LD focus on relationships that make sense to people will be effective for LLMs. Devon: Example of person who experimented with a corpus of Susan B. Anthony’s letters to create vector embeddings and see how they worked with search. Good example of how it could actually work.
Enhancing Discovery and Workflows: The Alma and Primo Linked Data Expedition in a Hybrid Metadata World session PDF presentation and recording Overview of achievements for supporting linked data in Alma/Primo LSP Overview of how institutions use BIBFRAME already, noting that learning is the main component in all current efforts On the horizon: Organization & location in addition to person entities, as well as additional languages for all entities, coming in H2 2025. Linked Open Data Form Editor general availability H1 2026
Our discussion: Prompted discussion about how Bibframe data is stored in Alma. Why would you create Bibframe data in Alma, if it isn’t being stored as triples? Does Alma just generate triples on the fly? Does Alma store Bibframe data? When you import BF data, how does it work?
Arches (Getty) similarly isn’t a triple store – it breaks the triples up into a SQL database. Same with the systems that power the Getty’s linked data service, and Islandora as well. Wikidata uses Wikibase – main data storage is a SQL database, but it also has a triple store. Pipeline converts SQL data into RDF data. Data is always created as SQL data, but kept in sync with the datastore. In some ways, it would be easier if we all just used the same technology! If we all just used Wikidata, even if it’s imperfect. One challenge with Wikidata is that there isn’t an ontology – the same type of items aren’t modeled the same way. Darnell Melvin at UNLV has worked on a pipeline for normalizing Wikidata data for use in local systems. PCC EMCO is kind of an effort towards that – how to use a single tool (or a suite of tools) in a standard way. “Organized chaos.”
Elliot attended Wikidata Workflows for ORCID and OpenRefine, Binghamton University lightning talk, PDF presentation and recording |
Topics for ongoing discussion | | Entity reconciliation for place names and venues updates on Ransom Center efforts? updates on UTL efforts to reconcile terms to PANA, or work with UCLA’s HAPI project?
Partnering with UTL Research Data Services on Linked Open Data efforts Wikidata management over time (per presentation and discussion last meeting) Paloma interested in researching Unique Resource Identifiers for Archival Resources Hanlin shared updates about her Wikidata research Project Funding for a year and a half – in very early stages now o Southern Archives project (Hannah, Katie, Devon)
|
