Without Heading?: Automatic Creation of a Linked Subject System

Λεπτομέρειες βιβλιογραφικής εγγραφής
Τίτλος: Without Heading?: Automatic Creation of a Linked Subject System
Συγγραφείς: Malmsten, Martin, Lundborg, Viktoria, Fano, Elena, Haffenden, Chris, 1982, Klingwall, Fredrik, Kurtz, Robin, Lindström, Niklas, Rekathati, Faton, Börjeson, Love
Πηγή: New Horizons in Artificial Intelligence in Libraries IFLA Publications. :179-198
Θεματικοί όροι: Subject headings, Knowledge organization in subject areas, Machine learning, Linked data, Information Studies, Biblioteks- och informationsvetenskap, Artificial Intelligence, Artificiell intelligens
Περιγραφή: Can problems with library subject headings systems designed and operated by humans be mitigated by machine learning? This chapter discusses work undertaken at Kungliga biblioteket/National Library of Sweden (KB) to explore the use of an automated system for identifying subjects that would sidestep the need for fixed headings from a controlled vocabulary. The new approach taken combined the natural language processing (NLP) capacities of large language models (LLM) with topic modelling, to cluster and order texts according to topics derived from the material, rather than any pre-defined terms reflecting existing biases. The chapter explains how BERTopic was tested on a text corpus compiled from KB’s digital collections and presents and analyses the results. The value of such critical exploration is to illuminate the shortcomings of existing systems that might otherwise remain unnoticed.
Περιγραφή αρχείου: electronic
Σύνδεσμος πρόσβασης: https://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-540448
https://doi.org/10.1515/9783111336435-014
Βάση Δεδομένων: SwePub
Περιγραφή
ISSN:03446891
DOI:10.1515/9783111336435-014