Heritage connector: A machine learning framework for building linked open data from museum collections

Abstract

As with almost all data, museum collection catalogues are largely unstructured, variable in consistency and overwhelmingly composed of thin records. The form of these catalogues means that the potential for new forms of research, access and scholarly enquiry that range across multiple collections and related datasets remains dormant. In the project Heritage Connector: Transforming text into data to extract meaning and make connections, we are applying a battery of digital techniques to connect similar, identical and related objects within and across collections and other publications. In this article, we describe a framework to create a Linked Open Data knowledge graph from digital museum catalogues, perform record linkage to Wikidata, and add new entities to this graph from textual catalogue record descriptions (information retrieval). We focus on the use of machine learning to create these links at scale with a small amount of labelled data, and models which are small enough to run inference on datasets the size of museum collections on a mid‐range laptop or a small cloud virtual machine. Our method for record linkage against Wikidata achieves 85%+ precision with the Science Museum Group (SMG) collection, and our method for information retrieval is shown to improve NER performance compared with pretrained models on the SMG collection with no labelled training data. We publish open‐source software providing tools to perform these tasks.

Relationship Prediction in a Knowledge Graph Embedding Model of the Illicit Antiquities Trade

“…We intend to use these statements, and the embedding model we derive from them, in a further study to create an automated relationship extraction pipeline (at present, the bottleneck is in the annotation and automatic extraction of relationships from unstructured text). We could then use the pipeline on other germane texts such as newspaper articles, the Panama Papers, judicial documents, and open museum collections (for an allied approach in terms of cultural heritage more generally, see Dutia and Stack 2021). Hardy's ongoing explorations of metal-detecting websites and other hidden-in-plain-sight fora (Hardy 2021) might also be amenable.…”

Challenges and opportunities in applying AI to evolutionary morphology

“…However, museum collections are rarely fully catalogued and even then it is difficult to search for a specific specimen or representatives of specific groups. This difficulty is because data is often inconsistent in quality and structure, particularly in large collections (Dutia and Stack, 2021). AI can play a key role in this, particularly when it comes to tasks of identifying, cataloguing, and locating specimens within collections.…”

“…has not yet been adopted on a large enough scale to allow searching global natural history collections and connecting specimens. Dutia and Stack (2021) recommend 'Heritage Connector', a framework and software for using ML to allow better connecting specimens in collections and publications. This software achieved a precision score of greater than 85% with science museum group records.…”

An adaptive method based on fuzzy logic and AdaBoost for exhibition of museum collections

“…Bowen et al [1] discussed the possible direction of the museum in the rapid development of digitalization. Dutia et al [2] used machine learning algorithms to identify Wikipedia entries related to the collection in order to improve the academic research efficiency of the collection. Authors in [3] utilized the convolutional neural network (CNN) model to identify and classify image collections, and the accuracy of collection online search has been improved.…”