[Ohiodig] Webinar from FromThePage: LLMs for Transcription and Metadata Creation

[Carleton, Janet (she/her)]

May be of interest. Free webinar.


Snipped from FromThePage<http://www.fromthepage.com/>'s Oct 4 newsletter "Transcribe New Collections Vol. 24"

LLMs for Transcription and Metadata Creation
[https://primary.proxybybento.com/content/7cca085cfa4e41fd85b70a24e0f3df60/lib/pluginId_7cca085cfa4e41fd85b70a24e0f3df60_75918/nov_2025_webinar.png]<https://kitsune.metrics.sentbybento.com/proxy/tracking/emails/VCf2VUtICLOhiZmvJV1GkvfvFHBM0NOW/click?signature=4f3080649c1657c24871d5d9825e36ac9f049db3&url=https%3A%2F%2Fcontent.fromthepage.com%2Fnov-2025-webinar%2F%3Futm_source%3Dbento%26utm_medium%3Demail%26utm_campaign%3Dbroadcast%26bento_uuid%3D0b261791-40dd-4128-a292-f4112efaa79b>

By: Willem Borkgren, Hannah Moutran, Katie Pierce Meyer, Josh Conrad, Devon Murphy & Karina Sánchez

Collections of digitized materials are growing faster than metadata can be added; can Large Language Models (LLMs) be one solution for consistent, quality metadata? The AI Metadata Creation Project at the University of Texas at Austin aimed to explore whether LLMs can offer a solution for improving the efficiency of developing metadata for large, digitized collections. Using a collection of digitized architectural publications, the group tested seven different models, using both APIs and web interfaces to test their ability in generating metadata, specifically subject headings, summaries of content, and named entities, as well as efficacy in metadata preparation tasks like Optical Character Recognition (OCR) of digitized materials. Our presentation will examine our preliminary findings as well as our testing workflow, so that it can serve as a model for others wishing to evaluate LLMs.

Leveraging expertise in metadata, collections, and technical tools, project members evaluated which tools to apply to the testing, created standardized testing criteria and a grading system, and met frequently to evaluate the results. Members are currently finalizing recommendations via a report, which investigates the efficiency of the LLMs in terms of results, cost, and time.

As this project nears completion, preliminary results suggest several models can sufficiently perform metadata creation tasks, but still lack accuracy in certain aspects such as subject headings and handling offensive language. Responsible, cautious, and targeted implementation of LLMs could greatly improve access to libraries' digital collections.

The webinar is on November 6, 2025 at 12:00 PM EDT, 11:00 AM CDT, and 9:00 AM PDT. Signing up will send you an invitation with the details and a follow up with the recording.
Sign Up Here<https://kitsune.metrics.sentbybento.com/proxy/tracking/emails/VCf2VUtICLOhiZmvJV1GkvfvFHBM0NOW/click?signature=4f3080649c1657c24871d5d9825e36ac9f049db3&url=https%3A%2F%2Fcontent.fromthepage.com%2Fnov-2025-webinar%2F%3Futm_source%3Dbento%26utm_medium%3Demail%26utm_campaign%3Dbroadcast%26bento_uuid%3D0b261791-40dd-4128-a292-f4112efaa79b>
[Fb]<https://www.facebook.com/FromThePageTranscription/>
[Ig]<https://kitsune.metrics.sentbybento.com/proxy/tracking/emails/VCf2VUtICLOhiZmvJV1GkvfvFHBM0NOW/click?signature=0d9234e2d63b273f207ecff277c8c1ac50379bf7&url=https%3A%2F%2Fwww.instagram.com%2F_fromthepage_%2F%3Futm_source%3Dbento%26utm_medium%3Demail%26utm_campaign%3Dbroadcast%26bento_uuid%3D0b261791-40dd-4128-a292-f4112efaa79b>
[Yt]<https://kitsune.metrics.sentbybento.com/proxy/tracking/emails/VCf2VUtICLOhiZmvJV1GkvfvFHBM0NOW/click?signature=7faa813faae9355d4e4bdcfe5f32903abb82d970&url=https%3A%2F%2Fwww.youtube.com%2F%40fromthepage%3Futm_source%3Dbento%26utm_medium%3Demail%26utm_campaign%3Dbroadcast%26bento_uuid%3D0b261791-40dd-4128-a292-f4112efaa79b>

Copyright © 2023 Brumfield Labs, LLC, All rights reserved.

You're receiving this newsletter because you signed up to transcribe documents at www.fromthepage.com<http://www.fromthepage.com/>.

Our mailing address is:
FromThePage
8606 Primrose Lane
Austin, TX 78757

---
Janet Carleton | Digital Initiatives Coordinator | Digital Initiatives | Mahn Center for Archives and Special Collections, Preservation & Digital Initiatives | OHIO University Libraries | Alden 333 | Athens, Ohio | 740.597.2527 | she/her | carleton at ohio.edu<mailto:carleton at ohio.edu> | Digital Archives - https://media.library.ohio.edu<https://media.library.ohio.edu/> | Archives Finding Aids - https://archivesspace.ohio.edu/

Visit our Socials
https://sites.ohio.edu/library-archives-blog/ | https://bsky.app/profile/aldenlibdigital.bsky.social | https://www.instagram.com/archives.ohiouniversity | http://pinterest.com/OhioDigiArchive/ | https://bit.ly/ou-uaSpotify | http://bit.ly/YouTube-OU-hist-films & http://bit.ly/ou-di-youtube

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.library.ohio.gov/pipermail/ohiodig/attachments/20251006/90b9fc7e/attachment.htm>