The potential and pitfalls of machine learning in the Congruence Engine context

Unwin, Jamie; Stack, John

The potential and pitfalls of machine learning in the Congruence Engine context

Download

Name

2024-06-11_ThePotentialAndPitfallsOfMachineLearningInTheCongruenceEngineContext_FINAL.pdf

Description

visibility:open

Size

2.48 MB

Format

Adobe PDF

Checksum (CRC64NVME)

39/yuj/u4Tc=

Resource type

Journal article

Creator (person)

Unwin, Jamie

Stack, John

Date published

January 25, 2023

Abstract

This article considers the role of machine learning (ML) in the Congruence Engine project. The authors bring their digital and ML experience to the project and here reflect on existing tools and approaches, the particular challenges of Congruence Engine endeavour, and possible solutions within and beyond the project. Although these ideas will develop as the project progresses, the authors draw on knowledge of the existing ML landscape and current digital collections practice as well as learnings from Heritage Connector, the Science Museum Group’s previous project in the same funding scheme as Congruence Engine. The authors propose that while significant advances in ML and the availability of open datasets such as Wikidata offer huge opportunities for linking heritage collections, this will require a pipeline model with iterative stages of human intervention. Closer relationships will need to be developed between human curators, researchers and users of ML and the technology and processes it requires and this article points to the likely areas of collaboration that Congruence Engine will explore and test.