Streamlining Metadata Creation: Implementing and Assessing AI Workflows to Improve Discoverability

Abstract

Anthologies of art song have often posed challenges to discovery as contents notes are not always adequately transcribed, making it difficult for users to know what songs are contained in each score. Transcribing contents notes can be difficult, especially when the songs are in multiple languages. Through a practical and real-world example, this paper demonstrates the application of automation and artificial intelligence to enhance cataloguing records with improved contents notes and evaluates the results through a user-centred lens. We highlight possibilities for this evolving technology as well as the challenges that it can pose and explore the concept of a cost-benefit analysis of metadata work with the element of artificial intelligence being considered in a holistic manner.