Medieval studies is a large field of research, covering the entire history and culture of Western Europe for more than a thousand years, from the fifth to the fifteenth centuries. The major conferences, like the annual International Medieval Congress in Leeds, contain more than 750 sessions with over 2,500 participants. There is a vast literature of printed and electronic publications, and an increasingly complex environment of digital resources of many different kinds. In this digital environment, however, there is nothing resembling the recent initiatives in closely related research fields (especially classics and ancient history, and early modern studies) to develop tools and resources for data linkage. At the heart of these initiatives is Linked Open Data – both as a paradigm and as a set of specifications, tools, and methodologies, loosely coordinated by the Linked Pasts Network. [1]

In this paper, we examine the current situation in medieval studies and present proposals for future directions in the application of Linked Open Data to this research field. Our proposals build on work already being carried out as part of the “Mapping Manuscript Migrations” (MMM) project, an international collaboration between the University of Oxford, the University of Pennsylvania, Aalto University, and the Institut de recherche et d’histoire des textes. [2] Funded by the Trans-Atlantic Platform under its Digging into Data Challenge for 2017-2020, Mapping Manuscript Migrations focuses on aggregating data from four major data sources relating to medieval manuscripts: the Schoenberg Database of Manuscripts, Medieval Manuscripts in Oxford Libraries, Bibale, and Medium. The methodologies being deployed include a unified data model derived from the CIDOC-CRM and FRBROO ontologies, and a matching and reconciliation process based on Linked Data identifiers and vocabularies.

For the MMM project, the main ontological classes are Manuscripts, Persons, Organizations, Places, and Works. While the Manuscripts class could be expanded to cover Objects more broadly, manuscripts are the most common surviving type of object from the medieval period and the most valuable evidence from a range of interdisciplinary research perspectives. As well as being the main focus of medieval research, they exemplify most of the issues connected with the deployment of Linked Data ontologies and vocabularies.

Each of the source datasets involved in the MMM project have their own preconditions and goals, and thus follow their own data modeling conventions. The Schoenberg Database records primarily provenance events and observations of manuscripts at specific points in time, based on, e.g., auction catalogs, while in Bibale and the Bodleian catalogue the main focus is on describing manuscripts as objects. In order to aggregate the datasets and use them with shared user interfaces and tools for answering a wide range of research questions, a unified data model for harmonizing them is needed.

The model applied in the project is based on the CIDOC-CRM and FRBROO ontologies, where the essential distinction between works, expressions, manifestations, and items are considered [3]. The most essential core classes in the model are (1) physical manuscript objects with properties such as shelf-mark, owner and title, and (2) provenance events that describe events related to the manuscripts, such as production, appearance in an auction, acquisition, etc. These ontologies were chosen as the basis because they are modern standards for data harmonization in museums and libraries, respectively, and they support event-based modeling needed in modeling provenance data that are essentially chains of events concerning the manuscript objects.

The CIDOC-CRM and FRBROO ontologies have been extended in the MMM project to suit the modeling needs of the source databases. Extensions include 1) subclasses of crm:E54_Dimension for representing the height, width, and folio, column, and line count of the manuscript; 2) specific properties for linking the entities to source databases, such as to manuscript records, observation entries, and data providers; 3) subproperties of crm:P1_is_identified_by to express different shelfmark conventions; and 4) representation of the roles of participants in events, such as author, scribe, artist, binder, patron or selling agent.

The majority of places in the source datasets for the MMM project can be matched to identifiers from one or more of the most widely used generic Linked Open Data vocabularies: the Getty Thesaurus of Geographical Names (TGN) and GeoNames. TGN is much smaller than GeoNames but more historically oriented and actively curated. Neither has anything like the coverage of the gazetteers produced by mapping agencies. For Great Britain, for example, TGN has only half as many places names as GeoNames, but this in its turn has only a fraction of the names provided by the Ordnance Survey. [4]

The MMM datasets, in which medieval, early modern, and modern place names are all present, have their preferred default vocabulary: GeoNames for Bibale, and TGN for Schoenberg and Oxford. None of these datasets makes use of a gazetteer specifically designed to cover medieval places and their medieval names. This is partly, at least, because the evidence in the MMM source datasets focuses on the history of manuscripts in the post-medieval world. The places referred to are usually (though not always) the same as places which exist today, and they are usually presented in modern languages, not in Latin. For medieval places in England, the best source of data is the Survey of English Place Names, published by the English Place Name Society and digitized as part of the Historical Gazetteer of England’s Place-Names. This is not yet available as Linked Data, though its creators claimed that it would be “relatively easy to convert it to RDF”. [5]

In the source datasets for the MMM project, “person” entities can cover everything from medieval authors to early modern owners to modern buyers and sellers of manuscripts. The three datasets are consistent in using VIAF identifiers as their primary source for persons. In the Schoenberg Database of Manuscripts, VIAF is the preferred source. In the Bodleian catalogue, VIAF identifiers are supplemented by other sources, including the Library of Congress, Wikidata, Sudoc, Gemeinsame Normadatei (GND), and the Oxford Dictionary of National Biography. Bibale also prefers VIAF, but adds identifiers from Wikidata and the ARK identifiers from the Bibliothèque nationale de France; Biblissima identifiers are also referenced.

None of the MMM datasets makes reference to any specifically medieval name authorities. There are many lists and catalogues of medieval persons and their names, and a significant number of these are available in digital form. [6] It is worth noting that several of the digital sources are commercial products, distributed by publishers like Brepols. These include Europa Sacra (with data on 30,000 church prelates) and the author indexes to the Library of Latin Texts. Some specific initiatives are already under way, including the work by the “Medieval Publishing” project at the University of Helsinki to transform Richard Sharpe’s into database form. [7] Publishing this dataset as Linked Data is envisaged as a second stage for this project.

Organizations are not always distinguished from places in the source datasets for the MMM project. The interface to the Oxford manuscripts catalogue simply includes organizations in its browsable list of “places”. This is because the majority of these organizations are medieval religious houses of various kinds, which are usually known by a place name, such as “Abingdon, Benedictine abbey”. Behind the scenes, however, they are linked to the VIAF (Virtual International Authority File) vocabulary. Other identifiers are also given, including Wikidata, the Library of Congress, Sudoc, and ARK identifiers from the Bibliothèque nationale de France.

In the Schoenberg Database of Manuscripts, on the other hand, organizations are combined with persons into a “Names” file. These are both medieval and modern, and include a significant number of religious houses. A large proportion of the organizational names have been matched to VIAF identifiers. Bibale’s organizations consist mostly of libraries and other institutional owners of manuscripts (especially medieval and early modern religious houses). VIAF is again consistently used, accompanied by identifiers from the Bibliothèque nationale de France, Biblissima, and Wikidata.

Organizations as a class raise various questions of scope and definition in a medieval context. In medieval England, for instance, Anglo-Saxon administrative units like “hundreds” could be considered primarily as places, since they were defined as geographical areas within a county or shire. But the Norman system of government overlaid on top of this framework after 1066 was less conceptually clear; the “honours” and “baronies” of the great lords consisted of various estates scattered across the country, which were primarily organizational and administrative in nature. [8] A specialist Linked Open Data vocabulary for medieval organizations in England could be drawn from existing lists of baronies and religious houses, together with data relating to offices within the structure of royal government and to the administrative hierarchy of the Church.

The Works class is problematic in the source datasets for the MMM project. They contain very little in the way of links to external Linked Open Data identifiers for works, and some do not even distinguish “works” – or even titles – as a separate element in their data. The Schoenberg Database, for example, does not contain separate records for works or titles; they are simply recorded as part of the observation record. In the Oxford manuscript catalogue, on the other hand, the a provides a unique ID number for more than 12,000 works. It records the “uniform title” (author + work), the “manuscript title”, and the “normalised manuscript title”. There are more than 360 links to external identifiers – primarily Pinakes for Greek works and Mirabile for Latin works. Bibale distinguishes between “Oeuvre” (i.e., work) and the “Text/Edition” (i.e., manifestation). There are unique identifiers for both of these, but no links to external identifiers.

There are numerous lists and catalogues of medieval religious and literary writings, and many of these are available in some kind of digital form. [6] Among the issues to be tackled here are the and its derivatives (the focus of much medieval writing, especially in the earlier Middle Ages); the multiplicity of anonymous works – often without any kind of title and often identifiable only by their opening words (their “incipit”); and the way in which many medieval “works” actually consist of annotations, commentaries, and glosses on earlier works, as well as re-workings and abridgements of them. The work of the “Medieval Authorship” project is relevant here too, since Sharpe’s Handlist identifies and provides titles for the writings of medieval British and Irish authors.

Each of the MMM source datasets attempts to go beyond simply using the shelf-mark and repository for identifying individual manuscripts. The Bibale database contains links to another IRHT database, Medium, where 109,000 manuscripts have their own ARK identifiers. In the Bodleian catalogue, each manuscript has a unique ID number, but there is no authority list for manuscripts (unlike persons, places, organizations, and works). The creation of ARK identifiers for Oxford manuscripts is currently being investigated by the MMM project and the Bodleian Library. In the Schoenberg Database, each of the nearly 25,000 manuscript records has a unique URL, but they are not accompanied by links to any external identifiers or to an authoritative shelf-mark or title.

The need for a system of Linked Data identifiers for medieval manuscripts is now broadly accepted, and has been the subject of several meetings over the last two years. Under the auspices of the IRHT, a project has been established to develop and test a specification for an International Standard Manuscript Identifier or ISMI. [9] There is still considerable work to be done in defining the scope and purpose of ISMI, as well as in confirming and building the technical solution. Even when the ISMI registry is operational, it will still take a major effort to create and embed these identifiers in local databases for manuscripts with known and established current institutional locations and shelf-marks. Nevertheless, ISMI has the potential to revolutionize medieval manuscript studies by enabling large-scale inter-linking of heterogeneous manuscript data.

We believe that it is critical to develop, publish and re-use Linked Open Data ontologies and vocabularies in order to further the interoperability of digital resources for medieval studies. The combination of CIDOC-CRM and FRBROO ontologies, with some extensions, appears to function as a suitable data model for representing manuscripts. The more generalist vocabularies like VIAF and TGN already have a reasonably good coverage of medieval persons, places, and organizations, as the MMM project has been able to demonstrate. But the level of that coverage has been inflated to some extent by the scope of datasets relevant to manuscript research, since they include many early modern and modern names as well as medieval ones.

The next priority for medieval studies should be the conversion of existing specialist vocabularies to Linked Data formats, accompanied by an investigation of ways in which they can be associated with the more general vocabularies. The work being done by the “Medieval Publishing” project provides a possible model here. While persons, places, and organizations might be considered the “low-hanging fruit” in this context, it will be more important and valuable for the longer term to tackle the more complex issues around works and manuscripts – the latter within the framework envisaged by the ISMI initiative.