IGeLU 2025

Conference

Over the past year, The University of Haifa's Younes and Soraya Nazarian Library had several initiatives involving AI for metadata extraction. In this presentation I will outline three methods we developed so far:

1. A project by two graduate students involving data extraction via LLM models and R studio. Their model is focused on data extraction from a large collection of periodicals and uses transliteration of spoken words, data mapping and analysis, and finally exporting all data to an Excel file that can be loaded into Alma
2. Alma cataloguing AI Assistant – we examined both creating and enhancing bibliographical records capabilities to arrive at the most efficient workflow.
3. OCR for books in non-Latin scripts (Hebrew and Arabic) 

Conclusions:

AI can help streamline workflows, especially when dealing with large quantities of data

Human in the loop – specifically for books in non-Latin scripts

Thinking outside of the box results in more efficient workflows 

Practical Applications