Exploring a large graph of historical objects: the Mapping Manuscript Migrations knowledge graph

The Mapping Manuscript Migrations (MMM) project took data from three disparate sources relating to the history of Western medieval and renaissance manuscripts, and transformed them from TEI-XML documents or relational databases into a knowledge graph. This graph consists of more than 24 million RDF triples, together with a data model drawing on ontologies like CIDOC-CRM and FRBROO, and extensive vocabularies harmonized across the data sources. The graph includes more than 222,000 manuscripts, 56,000 people and organizations, and 5,000 places, as well as 937,000 events and 435,000 works. This paper examines and evaluates methods for traversing this graph, and for exploring, analysing, and visualizing the data.

MMM provides a public portal, with the capability to browse, search, filter, and download the data in the form of lists and CSV files. The Sampo-UI software used for the portal can also create map-based visualizations of the data to show the locations of manuscript production, together with last-known locations and the historical trajectories of manuscript ownership. The entire data package can be downloaded from the Zenodo data repository or inspected through the Linked Data Finland endpoint. This endpoint can also be searched with SPARQL queries. The data can be imported into other software environments like nodegoat or ResearchSpace, which enable different kinds of exploration and visualization.

Using a set of specific research questions drawn up during and after the MMM project, the paper will examine the affordances of these different methods for exploring the knowledge graph. The use of SPARQL queries, in particular, is a valuable way of assessing the types of analysis and visualization which can be usefully applied to the graph, as well as for diagnostic exploration of the contents of the source data as reflected in the RDF transformations. One of the lessons learned from this process has been the extent to which the scope of the source datasets can impose limitations on quantitative and descriptive queries across the graph.

SPARQL can also demonstrate the ways in which lacunae in the MMM data (in such areas as gender, occupation, and other biographical details) can be compensated for by drawing in this kind of information from other linked data sources. Being able to analyse and visualize data across multiple knowledge graphs, connected through URIs, is a major additional benefit of the Linked Open Data approach taken by projects like MMM.

References
Burrows, T, Emery, D, Fraas, M, Hyvönen, E, Ikkala, E, Koho, M, Lewis, D, Morrison, A, Page, K, Ransom, L, Thomson, E, Tuominen, J, Velios, A and Wijsman, H “Mapping Manuscript Migrations Knowledge Graph: Data for Tracing the History and Provenance of Medieval and Renaissance Manuscripts”, Journal of Open Humanities Data, 6:3 (2020). DOI: https://doi.org/10.5334/johd.14
Mapping Manuscript Migrations: portal https://mappingmanuscriptmigrations.org/en/
Mapping Manuscript Migrations: Linked Data Finland https://www.ldf.fi/dataset/mmm
MMM SPARQL endpoint http://ldf.fi/mmm/sparql
Nodegoat http://nodegoat.net
ResearchSpace https://researchspace.org