LTS Newsletter: April 2025

Library Technology Services, Harvard Library

Welcome to Library Technology Service’s occasional newsletter for anyone interested in Harvard Library IT.

Harvard Analytics and Reporting Tool (HART) expands library reporting capabilities

HART is the reporting system for Harvard Library data sources such as Harvard Depository, ReCAP and HathiTrust. In September, 2021 Harvard Library was onboarded as the first tenant of Harvard’s enterprise Oracles Analytics Server. HART gives library staff the ability to report on Harvard Depository and ReCAP inventory in conjunction with bibliographic, holdings and item data updated from Alma on a daily basis. Reporting on every field in bibliographic and holdings records is available in HART. LTS continues to partner with the ATS-RAID team to add reporting data for Google Library Inventory including HathiTrust, and Aleph circulation history. Requirements gathering is underway to implement ArchivesSpace in HART. The deep expertise and commitment to users by ATS-RAID colleagues have advanced library reporting skills and contributed to the creation of enterprise dashboards for Electronic Theses and Dissertations, Data Sync, Cataloging Statistics and HD/ReCAP item tracking.

Visit the HART Library Reporting wiki page to request a HART login, learn more about available data sets, or request assistance with your reporting needs. ARWG holds monthly drop-in sessions for users of HART and Alma Analytics. All are welcome!

Reimagining Discovery

The Reimagining Discovery team completed the pipeline to ingest finding aids into Collections Explorer reached as well as deploying the newly designed UI to our development environment! Work included in this milestone:

Finalized data transformations including work to account for wide range of dates.

Finalized embedding model choice with the help of a cohort of librarians. Cohere Embedv3 multi-lingual model performed at 85-88% precision for our metadata.

Determined that we need to devise a better chunking strategy to ensure even better precision for our metadata.

Deployed the newly designed UI to development environment so that we can perform searches and see results.

QA environment deployed.

See a walk-through of the new UI in the first 5 minutes of our Demo for Sprint 21.

Coming in May:

Add metadata filters and update the LLM prompt to summarize results in the UI

Begin QA ingest with new chunking strategy

Start evaluation of LLMs for summarization and suggested searches; relevancy tuning of results; and security measures for production

ArchivesSpace Harvester Pipeline to Primo

The newly developed ArchivesSpace (ASpace) harvester pipeline is now in production! This updated pipeline streamlines the harvesting of finding aids from ArchivesSpace for enhanced discovery through Primo (HOLLIS for Archival Discovery) and LibraryCloud. Built using the same robust pipeline framework that is used by JSTOR Forum, the new ASpace harvester allows us to retire the legacy pipeline and ensures a more reliable and maintainable process moving forward. All development and testing were successfully completed during the recent sprint, and EADs are now actively flowing into LibraryCloud and Primo through the new pipeline.

New Feature in Alma: AI Metadata Assistant

A new tool is available in the Alma Metadata Editor that uses AI to create metadata based on images you can upload into the MDE. The tool can be used to create or enhance a bib record in Alma. The tool can facilitate cataloging by doing data transcription, such as for Tables of Contents, and the ‘assistant’ will (mostly) put information into the correct fields. All records need careful review as the tool does not replace the intellectual work needed for cataloging. More information is available in the Alma Staff Documentation wiki.