I wrote a piece on one of AI’s biggest challenges (model collapse) and how the solution to that is just another Wednesday for librarians.
#libraries #librarytech #AI #cataloging
www.linkedin.com/posts/stuart...
Whilst working on visualizations for my upcoming post on how librarians have already solved one of AI's biggest problems nearly 150 years ago, this one was generated too. Not suitable for my post, but it was too good not to use somewhere else.
#AI #GLAM #libraries #librarytech
Full methodology, all discoveries, complete tech stack documented here (LinkedIn article, sorry, but it's where I published it):
www.linkedin.com/pulse/follow...
Methodology works for any adapted literary work, museum collection or cultural dataset.
#DigitalHumanities #KnowledgeGraphs #LinkedData
Also built the wrong graph first (casting database instead of thematic analysis) and had AI generate completely incorrect Wikidata QIDs.
My Scarecrow: a river in Italy 🤦
My Wicked Witch: a German municipality
Verification matters.
Built it with free tools: Wikidata (SPARQL), VIAF/ISNI, TMDB, Neo4j.
Used Claude to help write Python scripts (I can't code).
The secret: Persistent Identifiers. When your data sources have PIDs, everything connects almost magically.
The graph revealed patterns I'd never have found through traditional reading:
• Dorothy in The Wiz (1978) is a 24-year-old teacher, not a child—reflects post-Civil Rights focus on adult self-determination vs. childhood nostalgia
• "Good vs Evil" completely inverts from 1939 to Wicked 2024
I spent a weekend building a knowledge graph of The Wizard of Oz universe (779 entities, 338 relationships) to understand how stories transform across 125 years.
Turns out L. Frank Baum's mother-in-law was a radical suffragist who influenced Oz's feminist themes—20 years before women's suffrage. 🧵
🎵Just you and I identifyyyying graph-ity🎵
Working on a little something, stay tuned!
#knowledgegraphs #linkeddata #oz #wicked #frankbaum #thewonderfulwizardofowl #needmoreozandlinkeddatapuns #glam
Tomorrow I’ve got an article on funding academia vs corporations. Visit my LinkedIn tomorrow!
Why generating an image like this one is extremely tough (and why this image generation is still a failure) on an article on my LinkedIn page tomorrow!
#AI #LLMs #GAI #Karate
There’s MARC21, Unimarc, Pica3, Pica+, Intermarc I know of - and these are widely used in Western Europe (excluding local variants of these variants), and I believe Eastern Europe has again its own Marc standards.
After having watched #kpopdemonhunters, if #Eurovision decides to block Israel next year, may I suggest entry for South Korea in their stead?
Does “Japan” (Nihon) essentially mean “book day”?
Still new to BlueSky - but looking for info, chats, people and news on #LinkedData #Libraries #SemanticWeb #KnowledgGraphs and a touch of #AI and #LLMs.
Connect with me if you haven't already!