Reranking Hits in Public Library Catalog Search with Learning to Rank
Michael Preminger, Henrik Holtvedt Andersen · Cataloging & Classification Quarterly · 2025
This article reports experiments with learning to rank (LTR) in the public library context. We used training labels based on human relevance judgments of query-document pairs, along with combinations of features drawn from catalog data, to train ranking models using two algorithms. We evaluated the models by measuring the ranking shifts of chosen hits. The results indicate that LTR is a method with much potential to improve the ranking of materials in a public library catalog.