This article is available at the URI http://dlib.nyu.edu/awdl/isaw/isaw-papers/20-6/ as part of the NYU Library's Ancient World Digital Library in partnership with the Institute for the Study of the Ancient World (ISAW). More information about ISAW Papers is available on the ISAW website.
©2021 Ethan Gruber and Andrew Meadows; text and images distributed under the terms of the Creative Commons Attribution 4.0 International (CC-BY) license.
This article can be downloaded as a single file
ISAW Papers 20.6 (2021)
Numismatics and Linked Open Data
Ethan Gruber, American Numismatic Society, and Andrew Meadows, New College, Oxford
In: Sarah E. Bond, Paul Dilley, and Ryan Horne, eds. 2021. Linked Open Data for the Ancient Mediterranean: Structures, Practices, Prospects. ISAW Papers 20.
URI: http://hdl.handle.net/2333.1/q83bkdqf
Abstract: This paper is a revision of “Coinage and Numismatic Methods. A Case Study of Linking a Discipline,” which was authored by Meadows and Gruber and published by New York University’s Institute for the Study of the Ancient World’s electronic article series, ISAW Papers, in 2014.1 This was one of thirty articles published in a special series dedicated to Linked Open Data methodologies for the ancient world following the 2012-2013 National Endowment for the Humanities-funded Institutes for Advanced Topics in the Digital Humanities program, Linked Ancient World Data Institute (LAWDI).
Library of Congress Subjects: Linked data; Numismatics, Ancient.
The Opportunity
As a type of evidence for the ancient world, coinage is unique. Coins are monetary objects, and thus a key element in modern attempts to reconstruct the workings of the ancient economy. For example, through the process of “die-study” it is possible to determine with some degree of accuracy how many dies were used to strike a given coinage. This provides us with a way to quantify ancient monetary production. Since coins can also be attributed to particular rulers or cities with some degree of certainty, this makes it possible to ascertain the monetary output of different cities, kingdoms and empires, and to compare them with one another. There now exists a substantial body of scholarship devoted to the estimation of size of production, but comparatively little as yet to its broad analysis or representation in interactive media such as timelines or maps.2
Coins are also archaeological objects in that they have findspots. Coins within archaeological contexts have much to tell excavators about the contexts they are digging, but also more broadly about the monetary profile of the site they are excavating compared to others of similar or different types; from multiple sites a regional history may emerge (see, e.g., Reece, 1982). But findspots also give coins a trajectory. If we know where a coin was made and where it was found, we have evidence for movement, connectivity, and economic circulation (see Map 1).
Few archaeological objects from antiquity can be mapped from source to deposition with such certainty as coins, and yet again we are only beginning to exploit the possibilities of this evidence in analytical and representational tools. Moreover, with the advent of the metal detector, individual coin finds and their recording are no longer confined to excavation material. The Portable Antiquities Scheme (PAS) in the United Kingdom, for example, now has recorded findspots for some 412,000 coins (Pett 2014; see Map 2). The PAS has led to new works of synthesis (e.g., Leins, 2012; Walton, 2012).
Other similar databases have seen online publication in recent years. Finds databases for Roman coins found outside the Roman Empire exist for both Germany and Poland. The University of Oxford is creating a large-scale database of Coin Hoards of the Roman Empire (CHRE), to join with the American Numismatic Society-led initiatives Coin Hoards of the Roman Republic (CHRR) and the Inventory of Greek Coin Hoards (IGCH) to paint a broader picture of coin hoarding spanning a millennium from Archaic Greece to the end of the Roman Empire.3 These projects are at varying stages of integration into the wider numismatic research ecosystem. Only when it becomes possible to compare data sets across multiple modern source countries will we be able to write the larger monetary history of ancient imperial spaces.
Unlike most other forms of archaeological evidence, coins are official objects: their designs and inscriptions can tell us about the intentions of their issuers and, perhaps, the preconceptions of their users (See e.g., Fig. 1). The iconographic and epigraphic repertoire of ancient coinage is a huge, and substantially un-mined resource for examining areas from local religion to imperial economic policy; from individual political ambition to communal statements of identity. And there is scope here, as recent work has shown (Kemmers, 2006; von Kaenel and Kemmers, 2009), to marry the evidence from findspots to that of the iconography of the objects, and thereby expose patterns of administration invisible from other sources.
(ANS 1944.100.38334, American Numismatic Society, accessed October 26, 2020, http://numismatics.org/collection/1944.100.38334.)
Image used under CC0 license.⬈
Finally, there is the sheer quantity of material that survives. As we have already noted, after 20 years of recording, the PAS contains information on 412,000 coins, and the rate of discovery, and thus growth of material is not yet showing signs of abatement. This is just one country. Initiatives exist to record individual finds as well as hoards in numerous other European countries, most recently, the Netherlands with the enactment of the Portable Antiquities of the Netherlands (PAN).4 Moreover, the collections of Museums—both local and national—raise the numbers of coins available for study literally into the millions (Callataÿ 1997b). To these sources we must add also those coins that appear in commerce every year, most of which are publicized online through various auction houses (e.g., Fig. 2).
Numismatic material thus presents an exciting set of opportunities to address questions unanswerable through other source material, and to do so with substantial quantities of data, a significant amount of which is already available online in a variety of forms.
The Problems
However, two of the the principal barriers to the exploitation of this data lie in its sheer volume and its location in a variety of institutional and non-institutional settings. Many, many coins are described online, but they are described in different languages, are hosted in different systems and in different formats, according to different standards, and with different aims (compare Figs. 3a-3d).
To add to the complexity, numismatics has its own way of describing numismatic objects, necessitated by and tailored to features specific to coinage. Coins, for example, have two sides, both of which must be described. They are both pictorial and textual. Physical characteristics such as material, weight, diameter, and the relationship of the heads to tails (obverse to reverse) directions (axis) can be recorded. Coins have denominational systems that vary with time and place, and those times and places may require systems of chronology and geography that vary from standard, modern formats (e.g., “154/3-153/2 BC”, “first quarter of the second century”, “Byzantium”, “Constantinople.”) There is also a vast array of information about individuals involved in the production of coinage whose names and titles may vary over time, and find no easy analogue in any other discipline: “Augustus”, “Octavian”, “Tresvir”, “moneyer”, “die-engraver”. In a further twist, it is often necessary to record not just the details of an individual coin, but also its relationship to others: the context of its circumstances of discovery, a taxonomic arrangement, a commercial transaction, or its current physical disposition.
Solutions: Creating a Linked Discipline
Coinage is thus a rich source for the study of the ancient world, and the study of Roman Imperial coinage in particular is well established in the print medium. Roman numismatists have, over the past century, divided their discipline into four discrete areas of study. The basic structure of Imperial coinage has been the focus of a type corpus known as Roman Imperial Coinage (RIC), printed in 10 volumes, providing a basic description of each of the 40,000+ recorded varieties of the coinage. It is the standard reference work for all who catalogue and publish Roman coins in any context. These contexts may be divided into three separate types, which have formed the subject of the other three areas of focus of numismatic study: collections, hoards, and individual finds. Roman coins exist in the hundreds of thousands in the major public collections across the world. Hoards (coins buried together in antiquity) are found today in astonishing numbers across the former territory of the Empire. Single finds are similarly common, both within archaeologically excavated contexts, where their scientific value is enhanced, or as chance or metal-detector finds. As Fig. 4 describes, all four of these areas of study are fundamentally interlinked since collections may contain hoards and single finds, and any coin from any context must be described by RIC type for it to be properly published and usable in historical synthesis.
In a very obvious sense, Roman numismatics was a prime candidate for the introduction of a Linked Data approach to the entire range of publications required by the discipline. In 2011 the American Numismatic Society began, in collaboration with a number of strategic partners, the process of developing the necessary infrastructure for the creation of linked Roman numismatic data. These technological principles have been expanded to include a significant portion of Hellenistic numismatics at the date of this publication, and collaborators in England, France, and Spain are working on pre-Roman Gallic and Iberian coinage.5
Vocabulary and Ontology
A key element of this infrastructure are the stable URIs required to describe numismatic concepts. As noted above, there are elements of coin-description peculiar to numismatics that require a tailored approach to the creation of a discipline-specific vocabulary. Here we were able to harness a project (http://nomisma.org) established by Sebastian Heath and Andrew Meadows in 2010 to provide stable digital representations of numismatic concepts in the form of http URIs that also provide access to reusable information about those concepts, along with links to other resources. This allows us to build a graph of Roman coin data that is linked within Roman numismatics by the use of discipline-specific terms such as denominations (e.g., http://nomisma.org/id/denarius) or mints (http://nomisma.org/id/lugdunum), but also allows us to join the broader graph of ancient world data through the use of common identifiers such as Pleiades URIs (e.g., https://pleiades.stoa.org/places/167717) for ancient places and the Getty Art & Architecture Thesaurus (AAT) for other categories of concepts (e.g., denominations such as denarius: http://vocab.getty.edu/aat/300037266). By this simple decision we were able to ensure, in theory at least, the integration of numismatic data within the field of Roman numismatics, but also the permeation of numismatic material into other fields of study. The launch In July 2013 of a revised Nomisma.org system, based on Apache Fuseki,6 with a SPARQL endpoint and improved APIs, has allowed us to integrate Nomisma IDs fully into a number of numismatic projects.
By forming a concordance between Nomisma.org-defined concepts and URIs in external information systems, it is possible to integrate numismatic data into the broader ancient world Linked Open Data cloud. For example, a coin minted in Lugdunum, as defined in Nomisma, is by extension produced in its Pleiades equivalent. The underlying machine-readable data in Nomisma facilitate the direct publication of the American Numismatic Society’s Greco-Roman materials to Pelagios (via the Peripleo web application) through the ANS digital collection, MANTIS (http://numismatics.org/search). Similarly, a Nomisma API, built on an underlying SPARQL query, enables content from Nomisma’s partners (both small and large institutions) to make their materials available in Pelagios. This is an especially valuable feature for partners who simply lack the technical personnel or expertise to build these data exports directly into their own database systems. Of the nearly 200,000 Greek and Roman coins available in the Nomisma.org SPARQL endpoint, more than 134,000 are available to Pelagios through this API.
Today, more than 5,000 intellectual concepts have been defined by Nomisma.org across a growing number of periods, cultures and categories. While Greco-Roman concepts were the first focus of the project, it has grown to include partial coverage of Islamic numismatics and Medieval and early Modern European concepts. Categories of terms range from materials and denominations to geographic identifiers, like mints and regions, to personal, corporate, or dynastic entities responsible for the minting of coinage. Nomisma is overseen by a steering committee, and the responsibility for creating and maintaining IDs has fallen to discipline-specific working groups, such as for Roman, Greek, and Medieval numismatics.
Nomisma underwent another significant architectural overhaul in 2014, including a migration of data into a more standards-compliant model following modern Semantic Web specifications. This process included the introduction of the Nomisma ontology, developed by Karsten Tolle, a computer scientist at the University of Frankfurt, in collaboration with numismatists and developers on the Nomisma steering committee.7 The Nomisma ontology is not designed around closed-world principles in linked data. Numismatic data should not be exclusively confined to the properties and classes defined within the ontology. Rather, concepts in Nomisma are modeled primarily in the Simple Knowledge Organization System (SKOS), an ontology designed for the expression of taxonomies and internal and external linkages between terms. The Nomisma ontology as it applies to concepts further refines our application of SKOS by defining specific classes of information. The mint of Rome is designated by the URI http://nomisma.org/ontology#Mint (shorted to nmo:Mint). A denarius is an nmo:Denomination. The thesaurus uses properties and classes from numerous ontologies, from Dublin Core Terms for titles and bibliographic references to the WGS84 geographic ontology for geospatial coordinates.8 Rather than inventing a new property or class for every possible type of information, the ontology is intended to fill gaps in existing ontologies and facilitate the organization of numismatic information in a way that reflects how numismatists themselves classify and query objects. Therefore, properties in the ontology correspond to classes: a coin or coin type from Rome carries the property nmo:hasMint linking to http://nomisma.org/id/rome, and nmo:hasDenomination linking to http://nomisma.org/id/denarius.
A Type Corpus
The existence of a full print corpus of Roman Imperial Coinage (RIC) provided a ready-made framework for the creation of an online type corpus. A project was established in 2011 jointly by the ANS and New York University’s Institute for the Study of the Ancient World (ISAW) to build an online adaptation of this resource on the principles of Linked Data. This project benefited from head-starts in two areas. First the resource for the creation of necessary URIs existed in the Nomisma.org project; second, the collection database of the ANS (http://numismatics.org/search) already contained the descriptive elements of approximately one-third of the c. 40,000 known types of Imperial coinage. From these two resources we were quickly able, with technical implementation by Gruber and data under the management of Dr. Gilles Bransbourg, to establish Online Coins of the Roman Empire (OCRE, http://numismatics.org/ocre), a type corpus of Roman coinage. Initially launched in July 2012 and with the types from Augustus to Hadrian, the project was completed in 2017 (although errors continue to be fixed), following a three-year grant from the National Endowment for the Humanities (NEH).9
As work progressed on the creation of type records within OCRE, we created the Nomisma.org URIs necessary for their description. The Linked Data approach we have taken to the creation of OCRE has a number of obvious payoffs. An attractive feature is that by providing alternative names for all Nomisma.org concepts in multiple languages, we have been able quickly and easily build a multilingual interface (compare Figs. 5a and 5b), derived from SKOS-defined preferred labels in RDF extracted real-time from Nomisma.org’s APIs. To date 17 languages are supported, including English, German, and French (among the more common languages among our user-base), but also Bulgarian, Polish, and Arabic, which do see significant usage. The advantages to the Linked Data approach run deeper, too. Where the Roman Imperial Coinage type corpus can do nothing but describe the types themselves and illustrate a single representative example of a selected few types, OCRE has the power to link to multiple examples of a given type from multiple contexts. There are now 31 partners and 117,570 total specimens from an international range of museum and archaeological databases aggregated by Nomisma.org and made accessible through OCRE.
The evolution of OCRE, between enhancement of Nomisma’s infrastructure and software development surrounding its publication platform, Numishare, opens the door to similar projects in the future.10 The American Numismatic Society and the broader numismatic community have published four subsequent Greek and Roman type corpora following the same technical principles. In early 2015, the ANS, in collaboration with colleagues at the British Museum, released Coinage of the Roman Republic Online (CRRO, http://numismatics.org/crro). Later that year, PELLA (http://numismatics.org/pella), a type corpus of the coinage of Alexander the Great, was released. In late 2017, the first iteration of Seleucid Coins Online (SCO, http://numismatics.org/sco) publicly launched, and the final version of this and the first phase of Ptolemaic Coins Online (http://numismatics.org/pco) were published in November-December 2018. These latter three projects fall under the umbrella of Hellenistic Royal Coinages, a project funded by the NEH in 2017, and will conclude by 2020.
Type Corpus to Hoards
In parallel with the development of OCRE, the ANS collaborated with Dr. Kris Lockyear of the University College London’s Institute of Archaeology to create (with implementation by Gruber) an online database of Roman Republican coin hoards (CHRR, http://numismatics.org/chrr).11 This is based on Dr. Lockyear’s personal research database (originally in MS Access; see Lockyear 2007), which is a substantially enlarged version of a print volume (Crawford 1969). Published in 2013 and consisting mainly of hoards of exclusively Roman Republican materials, approximately 10% of the hoards also contained early Augustan coin types (Lockyear and Gruber 2013). Initially CHRR drew on machine-readable data provided by OCRE’s APIs for Augustan types and Nomisma.org to extract Republican coin type data to facilitate geographic and statistical analyses, but now it extracts data from CRRO’s APIs for Republican data (with the Nomisma IDs having been deprecated following the publication of CRRO). Lockyear’s cataloging burden is now significantly lowered since the responsibility for long-term maintenance of Roman Republican and Imperial coin type data falls into the domain of the editors of those online corpora.
OCRE and CRROs’ data APIs enable CHRR to extract machine-readable data in real-time in order to render HTML pages (see e.g., Fig. 6a), generate GeoJSON and KML for maps and timelines (the distribution of types within a hoard over time and space), and facilitate a wide variety of quantitative analyses. Similarly, the Linked Open Data for these hoards are ingested into the Nomisma.org SPARQL endpoint, enabling the mapping of the distribution of findspots for specific types directly within OCRE and CRRO (see e.g., Fig. 6b). Taking this one step further, these find-spot data apply to any distinct Nomisma-defined concept connected to published typologies. Map 1, above, displays the geographic distribution of denarii by means of physical specimens linked to denarius types defined in OCRE and CRRO and their individual findspots or by Roman Republican coin hoards with type URIs linked to nm:denarius. The distribution of hoards will be greatly enhanced once Coin Hoards of the Roman Empire is ready for publication as LOD.
Type Corpus to Collections
There are also obvious benefits to establishing links from typological records published in OCRE and similar projects to individual specimens housed in public collections. The practice of assigning unique URIs to individual objects, now recommended by the International Council of Museums, and adopted by the majority of major collections, allows for stable connections to be built between those objects and typological records that describe them. So, for example, a coin type record can link to multiple instances of that coin type held in multiple collections or archaeological datasets (although in actuality, the link is from the specimen to type within the underlying LOD). Fig. 7, for example, shows the CRRO record for RRC 286/1, with references to 102 specimens of the type from 12 different collections, from large national collections such as the ANS and British Museum to small ones such as Rutgers University and the Fralin Museum of Art at the University of Virginia, as well as numerous coins registered in the PAS.
Through the ingestion into the Nomisma SPARQL endpoint of the specific characteristics, such as the weight or die axis of each specimen, it is possible to create tools to analyze the characteristics of individual types or of coins issued in particular places or by specified emperors (see Fig. 8a). Analysis can also be carried out on the generic characteristics of types using a similar set of set of filters. See, for example, Fig. 8b, which compares the frequency of the depiction of deities on the coinages of Augustus and Nero. OCRE, CRRO, and other individual type corpora enable these statistical visualizations among types within their own series, but we are beginning to build user interfaces in Nomisma.org itself to enable analyses more broadly across all periods and cultures of coinage. For example, it is possible to visualize the change in average weight of denarii between the late Roman Republic and early Empire, a question that cannot be asked of data and user interfaces within OCRE or CRRO exclusively.
Type Corpus to Coin Finds
Given the huge numbers of coins that are now being recorded by schemes across Europe, there is obviously huge potential to ingest the details of these coins into Nomisma.org for query and visualization in online type corpora. In this case it is not just the physical characteristics of the specimens concerned, but also their findspots which serve as the basis of enhanced analysis through mapping. Fortunately the price of admission to the Linked Data community for Roman numismatics is low. Existing projects, with, in some cases, well developed and longstanding databases of finds do not need to change their recording practices or software. So long as their data can be mapped to Nomisma and type corpus URIs when they are exposed to the Web, then the finds may enter the graph of numismatic data on the Web. The UK’s Portable Antiquities Scheme has been using Nomisma URIs since at least 2011. The Antike Fundmünzen Europa and Finds of Roman Coins in Poland projects were integrated into the numismatic linked data cloud in 2015 and 2017, respectively.
Moreover, the exchange of data is potentially a two-way street. These ANS-published type descriptions are fully and freely downloadable, and may serve to populate other databases, without the need for their creators to redo the work of cataloging that has already gone into the creation of the type record. Just as now few librarians catalogue books from scratch, so in the future there will be little need for the cataloguer of coins to generate new descriptions of coins long known.
Linked Open Data Architecture, Applied
How does Linked Data help in technical terms? The most efficient method for maintaining the relationships between coin types and their associated coins and hoards is with the use of an RDF database and SPARQL endpoint. Type corpora published in the Numishare platform (e.g., OCRE) directly query the endpoint to deliver some types of services. In other cases OCRE interacts with REST APIs offered by Nomisma (simple web service interfaces which conduct more complicated SPARQL queries in the background). The second, SPARQL-aware version of OCRE was released in October 2013.12 The RDF requires three components of data: first, RDF representations of coin types, second RDF descriptions of physical coins or coin hoards, and finally, the RDF data of concepts defined by Nomisma.org, which enables links to be made from typological attributes (e.g., http://nomisma.org/id/ar for silver coins or http://nomisma.org/id/augustus for those of Augustus) between types and coins/hoards. The RDF model conforms to the numismatic ontology and Metadata Application Profile established by Nomisma. It is relatively simple, especially compared with CIDOC-CRM, and lowers the barrier for participation in these type corpus projects. To date, more than 30 organizations are contributing coin specimens, types, finds, and/or hoard data to the Nomisma.org numismatic research ecosystem.
We can see the relationship between “ideal” type descriptions and individual specimens more clearly by looking at an example. The structure for RIC Augustus 1A is represented as an nmo:TypeSeriesItem defined by the URI http://numismatics.org/ocre/id/ric.1(2).aug.543A.13 A relationship is established by the nmo:hasTypeSeriesItem property in the RDF model, below, associating object number 18207296 in the Berlin Münzkabinett to Augustus 543A (note that the RDF prefixes have been removed):
<https://ikmk.smb.museum/object?id=18207296> rdf:type nmo:NumismaticObject ;
nmo:hasAxis 3^^xsd:integer ;
nmo:hasCollection nm:mk_berlin ;
nmo:hasObverse [
foaf:depiction <http://ikmk.smb.museum/mk-edit/images/n7/7474/vs_opt.jpg> ;
foaf:thumbnail <http://ikmk.smb.museum/mk-edit/images/n7/7474/vs_thumb.jpg>
] ;
nmo:hasReverse [
foaf:depiction <http://ikmk.smb.museum/mk-edit/images/n7/7474/rs_opt.jpg> ;
foaf:thumbnail <http://ikmk.smb.museum/mk-edit/images/n7/7474/rs_thumb.jpg>
] ;
nmo:hasTypeSeriesItem <http://numismatics.org/ocre/id/ric.1(2).aug.543A> ;
nmo:hasWeight 3.87^^xsd:decimal ;
dcterms:identifier "DE-MUS-814819/18207296" ;
dcterms:title "Röm. Republik: C. Iulius Caesar (Octavianus), vor 27 v. Chr."@de ;
void:inDataset <http://ikmk.smb.museum/>.
In summary, the coin defined by the URI https://ikmk.smb.museum/object?id=18207296 has measurements, a reference to the coin type URI of RIC Augustus 543A, URIs defining the collection and source dataset (for deleting/refreshing the export in Nomisma), and thumbnail and reference image URLs defined in both the level of the obverse and reverse of the coin. Additional hoard or findspot information may be added, as well as International Image Interoperability Framework (IIIF) service metadata (to be described in the section below). This simple model for translating numismatic metadata into Linked Open Data utilizes several Nomisma-defined ontology properties, but also implements Friend of a Friend (FOAF), Dublin Core Terms, and Vocabulary of Interlinked Datasets (VoID).14
Note that the RDF description of this object in the Berlin collection does not explicitly denote its material (silver), denomination (denarius), or other typological attributes. Like a relational database, the Nomisma LOD graph enables queries by these attributes by the association between the coin and the coin type by means of the nmo:hasTypeSeriesItem property. Practically speaking, how does this affect OCRE? Rather than updating records in OCRE when the collections of new partners become available, the RDF database stands apart, making it easier to update Apache Fuseki with new collections and manage changes or deletions in collections already contained in the system. Therefore, when Nomisma gains a new partner, these coin data and associated findspots, images, and measurements become available immediately upon ingestion into the endpoint. Using SPARQL, OCRE (through the Numishare application code) can query Fuseki to display images of coins associated with particular types in record or search results pages. SPARQL query results can be serialized into KML and GeoJSON directly and displayed in maps, making it possible to make use of findspots from online coin hoard catalogs. Using mathematical functions inherent to SPARQL, average weights of coin types or specific typologies (e.g., denarii of Augustus) can be delivered to OCRE directly and rendered in the form of charts and graphs (Fig. 8a) by the Javascript library, Highcharts, which is capable of interpreting this data into HTML5 graphics.15
Participating in the Numismatic Linked Open Data Ecosystem
The Nomisma project has maintained a documentation section on its website with basic details on how to model basic specimen data into RDF16 While we have yet to write a formal Metadata Application Profile that details each class and property in the Nomisma data model and their specific semantic meanings, the basic principles for contributing to the numismatic LOD cloud are fairly straightforward:
- According to the basic principles of the Semantic Web and recommendations for museum objects from ICOM, each coin must be available at a distinct URI. Ideally, the URI should follow the Clean URL pattern: there should be no portions of the URL that are distinct to particular software platforms (like a ‘.php’ or other programming-specific file extensions) or request parameters (variables passed in the URL following the ‘?’ character).17 An example of a Clean URI is a coin in the ANS collection: http://numismatics.org/collection/1944.100.51606. The accession number forms part of the URI, and the collection/ namespace is stable regardless of software platform (there have been three frameworks used for the online collection since 2010).
- At present, we are only ingesting physical specimens related to published type corpora, and we only have the capacity to ingest specimens whose type attribution is clear and completely certain. Therefore, each coin contributed to Nomisma.org should have one
nmo:hasTypeSeriesItemlinking to a coin type URI in OCRE, PELLA, etc. There are some exceptions to this rule, as there is overlap between Hellenistic type catalogs. Early Seleucid coins struck in the name of Alexander the Great may appear in both PELLA and Seleucid Coinage Online, so it is possible to have more than one URI, but in different corpora. - Every coin must have a human-readable title (although it can be automatically generated), but images, measurements, and findspots are optional.
The Nomisma Data Model
The How to Contribute Data page (http://nomisma.org/documentation/contribute) on Nomisma.org will contain the most up-to-date documentation on the Nomisma data models for contributing numismatic objects (coins, medals, tokens, etc.). Fig. 9 is an illustration of a typical model for a coin, although there are several additional properties that may optionally be applied. A full definition of the required and optional properties is as follows:
- Coin URI: encoded differently in various RDF serializations: XML, JSON-LD, Turtle, etc. [required]
- rdf:type: A physical specimen always bears the class
nmo:NumismaticObject. The particular object type is defined within the coin type RDF (e.g., http://nomisma.org/id/coin). [required] - nmo:hasTypeSeriesItem: The URI for a certain attribution of a coin type. There should be no more than one per type corpus. [required]
- dcterms:title: The title of the coin. A language is recommended, and there may be more than one title for more than one language (limit 1 per unique language). [required]
- dcterms:identifier: An identifier is required. This is an internal, institution-specific ID, such as an accession number. [required]
- void:inDataset: The URI of the parent collection database. This URI should be the same as the one defined in the required
void:Datasetmetadata. [required] - nmo:hasCollection: A Nomisma.org URI for the (usually) museum or library collection to which the object currently belongs. This should be ignored for research or finds datasets (like PAS) where the coin is part of larger database export, but the current collection is unknown or not part of Nomisma. [optional]
- nmo:hasAxis: Die axis, a number between 1 and 12 encoded as an integer value according to the XML Schema Datatypes (XSD;
xsd:integer). [optional] - nmo:hasDiameter: Diameter in millimeters of a round object as a an
xsd:decimal value(though it is usually rounded to the nearest whole number). If slightly oblong, use the longest measurement. [optional] - nmo:hasWeight: Weight of an object in grams, as an
xsd:decimal. [optional] - dcterms:isPartOf: A URI of a coin hoard which has been published into Nomisma. Geographic visualizations on coin type pages will extract coordinates and labels from the Nomisma SPARQL endpoint. At present, only IGCH, CHRR, and a small handful of Roman imperial coin hoards are available as LOD in Nomisma. Coin Hoards of the Roman Empire will be published into Nomisma eventually. [optional]
Other measurement properties such as depth (thickness), minimum and maximum diameters, height, and width are defined in the Nomisma ontology, although at present no contributors have used these properties.
Images
Photographs of coins are typically published on the web in one of two ways: 1) the obverse and reverse photographs are merged into a single digital image file or 2) the obverse and reverse images are separate digital files, each with distinct URLs. Many organizations publish their digital images at varying sizes or resolutions. It is most common for collections to publish thumbnail images as well as reference images that are intended to be viewed in the pages for individual object records (typically 400-600 pixels in a particular dimension).
There are two properties borrowed from FOAF to capture thumbnail and reference image URLs: foaf:thumbnail and foaf:depiction. The particular pixel dimensions of these images do not matter; they are scaled to a standard height and width on the “browse” or “coin type” record pages. When the images representing the obverse and reverse are combined into a single digital file, the foaf:thumbnail or foaf:depiction should be inserted as a property directly within the nmo:NumismaticObject RDF. When separated into obverse and reverse images, the nmo:NumismaticObject should include the nmo:hasObverse and nmo:hasReverse properties that point to a data object (which can be a blank node or a resolvable URI) that contains the FOAF properties. See Fig. 9 for an example. Most partners in Nomisma.org provide separate image files.
In January 2017, the Nomisma.org RDF data model and Numishare platform were extended to support the encoding of IIIF service metadata in accordance with specifications published by the Europeana Data Model (Isaac and Charles 2016). The model prescribes several new triples for expressing more information about the foaf:depiction URI: defining this URI as an edm:WebResource (in the Europeana Data Model ontology), defining a svcs:service linking to a URI for IIIF Image API, and a dcterms:isReferencedBy property that points to either the info.json that describes the image service (if the image represents either the obverse or reverse image, Fig. 10a) or the IIIF Presentation API (a URI to the JSON manifest)18 if the foaf:depiction is a combined obverse/reverse file (Fig. 10b).
<https://nrs.harvard.edu/urn-3:HUAM:COIN03850_dynmc> rdf:type edm:WebResource ;
dcterms:isReferencedBy <http://iiif.harvardartmuseums.org/manifests/object/186057> ;
svcs:has_service <https://ids.lib.harvard.edu/ids/iiif/18736279> .
<https://ids.lib.harvard.edu/ids/iiif/18736279> rdf:type svcs:Service ;
dcterms:conformsTo <http://iiif.io/api/image> ;
doap:implements <http://iiif.io/api/image/2/level1.json>
foaf:depiction obverse image (http://numismatics.org/collectionimages/19001949/1944/1944.100.85688.obv.width350.jpg) as a IIIF service.⬈
<https://nrs.harvard.edu/urn-3:HUAM:COIN03850_dynmc> rdf:type edm:WebResource ;
dcterms:isReferencedBy <http://iiif.harvardartmuseums.org/manifests/object/186057> ;
svcs:has_service <https://ids.lib.harvard.edu/ids/iiif/18736279> .
<https://ids.lib.harvard.edu/ids/iiif/18736279> rdf:type svcs:Service ;
dcterms:conformsTo <http://iiif.io/api/image> ;
doap:implements <http://iiif.io/api/image/2/level1.json>
foaf:depiction combined image (https://nrs.harvard.edu/urn-3:HUAM:COIN03850_dynmc) as a IIIF service.⬈Findspots
In the event that an individual findspot is known, the nmo:hasFindspot property may be applied within the nmo:NumismaticObject. This property can point to a URI or be a blank node. The data object defining the findspot itself should be a geo:SpatialThing in the WGS84 ontology. Ideally, there should be a foaf:name for the textual, human-readable label of the place and a geo:lat and geo:long for coordinates. It is recommended to point to the parent town or region where the findspot is located, by using dcterms:isPartOf to point to a Geonames.org URI that defines a modern place. However, if the findspot is a an ancient place (e.g., a coin found during an excavation in Ephesus), the Pleiades URI may be used.
<http://opencontext.org/subjects/FCE38E48-ECBF-4819-F7AD-C1D92F804058> a nmo:NumismaticObject ;
nmo:hasFindspot <http://opencontext.org/projects/3> .
<http://opencontext.org/projects/3> rdf:type geo:SpatialThing ;
geo:lat 37.322559 ;
geo:long 37.034902 ;
foaf:name "Domuztepe Excavations"
VoID Data Dump Metadata
In order to contribute a data dump to Nomisma, there must first be some basic RDF about the dataset, including title, publisher, and licensing or rights statements. These statements are encoded with Dublin Core Terms in a void:Dataset object that represents the URI for the online collection as a whole. For example, the numismatic collection of the Fralin Museum of Art at the University of Virginia is accessible at http://coins.lib.virginia.edu/. The following RDF metadata defines this dataset:
<http://coins.lib.virginia.edu/> rdf:type void:Dataset ;
dcterms:description "The Fralin Museum of Art at the University of Virginia numismatic collection contains about 600 coins of mainly Greco-Roman origin."@en;
dcterms:license <http://opendatacommons.org/licenses/odbl/> ;
dcterms:publisher "University of Virginia Library" ;
dcterms:title "The Fralin | UVa Art Museum Numismatic Collection"@en ;
void:dataDump <http://coins.lib.virginia.edu/nomisma.rdf> ;
void:uriSpace <http://coins.lib.virginia.edu/id/>
The dcterms:title and dcterms:description are required elements, and should have languages associated with them. Multiple titles and descriptions in other languages are allowed. The publisher should be the organization responsible for publishing the data (required), and the void:dataDump should point to the raw RDF data that contains of the the coins in the collection. At least one of dcterms:rights or dcterms:license must be contain in the void:Dataset, and it is possible to include both. The rights may be a literal rights statement or point to a URI in accordance with those defined on http://rightsstatements.org. The license must be a URI published by Creative Commons or the Open Data Commons. Note that both the rights and license apply to the machine-readable data, not the images. Image licenses are subject to the terms provided by each individual contributor to Nomisma.
Conclusions
This paper, a revision and enhancement of a work published in 2014 by the Institute for the Study of the Ancient World in its ISAW Papers series, illustrates a current “State of the Field” of digital numismatics. Since its original publication there have been numerous significant enhancements in the data models and ontologies, as well as the information architecture of the platform over the last four years. While the paper has focused on the implementation of Linked Open Data principles as applied to coin types, specimen collections, and find/hoard databases (and how to contribute more specimens to these growing aggregations), it has not touched on some other areas of recent or future advancement. The ANS has sought to integrate its library and archival holdings more seamlessly into the numismatic research data ecosystem, the topic of a publication at Computer Applications in Archaeology 2016 and the Oxford-Paris Alexander conference in 2017 (Gruber 2016 and 2018).
In the future, we will see the publication and network analysis of Greek monograms as part of the Hellenistic Royal Coinages project, and, relatedly, the application of IIIF methodologies for the annotation of images with the URIs for these monograms. Nomisma partners in Berlin and Sicily are working on new ways of iconographic classification through hierarchical concepts, moving away from the traditional text-based (and highly variant) mode of describing types. Colleagues in Frankfurt are beginning to experiment with machine learning and computer vision for identification and classification, made possible by rigorously curated, semantically structured numismatic Linked Open Data.
Appendix: Cleaning and Linking to Type Corpora with OpenRefine
Over the years, Gruber authored one-off PHP scripts in order to match text-based coin type reference numbers to URIs, particularly to link the collections of Berlin and the British Museum to the OCRE project. These scripts often required complicated string parsing and regular expressions and were far from perfect in making successful matches, although they were adequate for 80-90% of these collections. Each of these scripts oftentimes required several days worth of labor, and a better solution was sought to minimize the labor Gruber invested directly into this process. Data cleaning in OpenRefine began in 2016, and the advantage of this software platform is that curators at partner institutions can learn how to clean and link their own data to Nomisma and coin type URIs, thus (hopefully) removing Gruber completely from the Nomisma publication workflow.19 This section of the article may therefore be useful to the reader in order to understand the method by which data may be normalized to the Nomisma vocabulary with references to additional tutorials in order to develop hands-on expertise with this tool.
Initially developed by Google before being freely released to the open source community, OpenRefine is a software application that is used to clean large amounts of data. It has seen widespread usage throughout the cultural heritage sector to normalize data and reconcile entities (as strings) to their equivalent URI defined in a wide variety of LOD-oriented vocabulary systems. One of the most common use-cases is reconciling entities to Wikidata through a common (and well-documented) JSON-based API, but it is possible to follow this API documentation to build reconciliation services on other data systems.20 The are many tutorials focused on humanities data and researchers that may serve as guides for using OpenRefine and its reconciliation features.21 It is not the intent of the authors to write a tutorial here, but rather outline a strategy that will enable a museum specialist or archaeologist to link their spreadsheet data to coin types defined in OCRE or other corpora as efficiently as possible.
Nomisma.org Reconciliation API
The Nomisma.org Reconciliation API was launched in October 2017.22 It is built upon the Apache Solr search index that underlies the Nomisma architecture, and the API draws on the preferred and alternative labels in any language contained in the RDF for any Nomisma concept. It is therefore possible to reconcile spreadsheets that contained the name forms of denominations, people, mints, etc., in French, German, and other languages to the appropriate Nomisma URI.
Virtually any category of concept can be reconciled against the Nomisma API, but the most common types of of categories are mints, regions, denominations, materials, and people associated with the coinage (as authorities, issuers, or depicted in portraits). The Nomisma services support a wide range of the official APIs (detailed at http://nomisma.org/apis/reconcile), such as the Preview (a popup window that displays the definition of the concept, which is useful for disambiguating people and places that have the same or similar names) and the Autosuggest.
Already, we have used these APIs for our own data cleaning purposes: reconciling entities found within Text Encoding Initiative (TEI) ebooks in the American Numismatic Society Digital Library, to prepare spreadsheets of coin types for Seleucid and Ptolemaic Coins Online for publication, to clean up hoard data for the next phase of the digital Inventory of Greek Coin Hoards, and to preprocess spreadsheets of data about the physical specimens in the University of Graz collection for reconciliation directly against RIC types in OCRE. Although Gruber is the developer of many tools and visualization interfaces that are used by students and scholars of numismatics, archaeology, and classical studies, it is rare that he actually uses these tools for research. The OpenRefine Reconciliation API is a notable exception, and it has saved hundreds of hours of labor in the year since this feature has been developed when compared to the previous method of writing complicated one-off PHP scripts to generate Nomisma IDs and normalize large quantities of spreadsheet data (for example, from the Egyptian National Library collection23). It took roughly one week to develop this API, and was therefore well worth the investment.
Numishare Reconciliation for Coin Types
Directly after the publication of the Nomisma.org Reconciliation API, the extension of these features into the Numishare platform was undertaken making it possible to normalize type references against any corpus published in that framework, such as OCRE or CRRO. Like Nomisma, the reconciliation services built on top of Numishare are based primarily on creating an interface between OpenRefine and queries to a Solr search index. The “title” of the type (which contains the number) is indexed into Solr, and that is the primary field used for reconciliation. In cases like CRRO, PELLA, or Seleucid Coinage Online, where the corpus is published in a single, sequentially numbered volume, the matches are pretty straightforward. The title is simply "RRC 100/2" or "Price 100," so typically one would not need additional search properties to normalize a spreadsheet against RRC or Price numbers. OCRE is more complex, and it helps to understand how the titles have been structured. The title consists of the RIC volume number (according to our standard description), the ruler(s) according to the RIC section title, and the RIC number. The exceptions are RIC 6-8, where the section is based on the mint name (standard English preferred label from Nomisma) instead of authority.
Because RIC volume number descriptions from museum or archaeological database fields may vary, it is important to isolate the RIC number alone (excluding the volume number, page number, etc.) and use additional property columns to disambiguate the RIC number. For the most accurate matches across all volumes of RIC, the denomination, authority, and mint should be used as additional properties. Since these properties match only on the English-language documents indexed into Solr, the mint, authority, and denomination columns should have previously been reconciled to Nomisma concepts and normalized to the English preferred label from Nomisma (since this is what OCRE also uses). As a test case, about 90% of the University of Graz Roman imperial coins were successfully matched to OCRE URIs on the first three passes of RIC number + various combinations of property columns in just about five minutes. The remaining matches were made by comparing other data with the Preview API or the Autosuggest.
The Preview API constructs a pop-up window of typological information extracted from OCRE in real time (transformed from NUDS/XML into JSON), including an example specimen’s photographs, if available in the Nomisma.org SPARQL endpoint.24 After completing the University of Graz reconciliation, Gruber exported a concordance list of Graz and OCRE URIs in the form of CSV, and emailed it to Elisabeth Steiner, the developer responsible for the digital numismatic collection. She then integrated these OCRE URIs into their data and generated a Nomisma-compliant RDF export directly from their Fedora-based institutional repository. After loading the VoID RDF metadata from Graz in the Nomisma.org administrative panel, nearly 2,000 coins were imported into the Nomisma SPARQL endpoint and were immediately accessible in OCRE.
To-Do: Writing Nomisma RDF Directly from OpenRefine
The steps outlined above illustrate the process of importing spreadsheet data into OpenRefine, linking typological attributes to Nomisma.org SKOS concepts and coin type URIs defined in ANS-published online corpora, and then exporting better-quality and LOD-oriented spreadsheet data back out of OpenRefine. Technical expertise is still required to transform that better-quality CSV data into Nomisma RDF, either by writing code to generate the RDF from the database platform (as, for example, Dan Pett did for the Portable Antiquities Scheme, and Elisabeth Steiner did for Graz) or creating a script to transform the CSV data directly into RDF (which Gruber did numerous times for collections such as the British Museum or Metropolitan Museum of Art). The latter approach is not sustainable in the long-term. Our aim is to develop an extension to OpenRefine (perhaps by building onto an existing Linked Open Data plugin) that will allow users to output Nomisma-compliant RDF directly from the software without having an intermediary script to handle this phase of the publication workflow. Data literacy is (or should be) a fundamental building block in Digital Humanities pedagogy, and we are hopeful that universities will begin to offer OpenRefine tutorials to students and scholars alike across all scientific or humanities disciplines on a more consistent basis (as has become more common for other tools like R and GIS). This will enable future generations of scholars and curators to become more directly involved in the Nomisma.org data normalization and ingestion processes.
In an ideal world, software developers working on museum or archaeological databases would build export procedures that follow the open community data standards such as linked.art or Nomisma.org. However, an effective short-term solution might be to transform reconciled CSV into RDF and store it as a static file on a web server. In either case, these APIs make it possible for curators and other specialists to link their coins to coin type URIs published by the American Numismatic Society.
Works Cited
Callataÿ, F. (1997a). Recueil quantitatif des émissions monétaires hellénistiques. Wetteren: Editions Numismatique romaine.
______. (1997b). “Quelques estimations relatives au nombre de monniaes grecques: les collections publiques et privées, le commerce et les trésors,” RBN 143, pp. 21-94.
______. (2003). Recueil quantitatif des émissions monétaires archaïques et classiques. Wetteren: Editions Numismatique romaine.
______. (2005). “A quantitative survey of Hellenistic coinages: recent achievements” in Z. Archibald, J. K. Davies, and V. Gabrielsen. (2005). Making, moving and managing: The new world of ancient economies, 323-31 BC. Oxford: Oxbow Books, pp. 73-91.
______. (2011). Quantifying monetary supplies in Greco-Roman times. Bari: Edipuglia.
Crawford, M. H. (1969). Roman Republican coin hoards. London: Royal Numismatic Society.
Gruber, E. (2013). “Recent Advances in Roman Numismatics.” M.A. thesis, University of Virginia. DOI: 10.26608/gruber_roman_numismatics.
______. (2016). “Linked Open Data for Numismatic Library, Archive and Museum Integration,” Proceedings of the Computer Applications and Quantitative Methods in Archaeology Conference (Oslo), 29 March-2 April, 2016. Tübingen: University of Tübingen Library. DOI: 10.5281/ZENODO.1304269.
______. (2018). “Linked Open Data and Hellenistic Numismatics,” Proceedings of the Linked Open World: Alexander the Great, Transnational Heritage and the Semantic Web conference (Oxford), 2-4 April, 2017. Bordeaux: Ausonius Éditions. DOI: 10.5281/zenodo.1484529.
Isaac, A. and V. Charles (2016). “Guidelines for submitting IIIF resources for objects in EDM.” Europeana. Accessed at https://pro.europeana.eu/files/Europeana_Professional/Share_your_data/Technical_requirements/EDM_profiles/IIIFtoEDM_profile_042016.pdf.
Kemmers, F. (2006). Coins for a legion: An analysis of the coin finds from [the] Augustan legionary fortress and Flavian canabae legionis at Nijmegen. Mainz: Philipp von Zabern.
Leins, I. (2012). Numismatic data reconsidered: coin distributions and interpretation in studies of late Iron Age Britain. PhD Thesis, Newcastle University: http://hdl.handle.net/10443/1467.
Lockyear, K. (2007). Patterns and process in late Roman Republican coin hoards, 157-2 BC. Oxford: Archaeopress.
Lockyear, K. and E. Gruber (2013). “From Dbase III+ To The Semantic Web: Twenty-Five Years Of The Coin Hoards Of The Roman Republic Database,” Proceedings of the Computer Applications and Quantitative Methods in Archaeology Conference (Perth), 25-28 March, 2013. Amsterdam: Amsterdam University Press. DOI: 10.5281/ZENODO.1035892.
Meadows, A. and E. Gruber (2014). “Coinage and Numismatic Methods. A Case Study of Linking a Discipline.” ISAW Papers 7.15. Accessed at http://dlib.nyu.edu/awdl/isaw/isaw-papers/7/meadows-gruber.
Reece, R. (1982). “Economic history from Roman site-finds,” Proceedings of the Ninth International Congress of Numismatics (Berne), Vol. I, pp. 495-502.
von Kaenel, H.-M., and F. Kemmers. (2009). Coins in context I: New perspectives for the interpretation of coin finds: colloquium Frankfurt a.M., October 25-27, 2007. Mainz am Rhein: Philipp von Zabern.
Walton, P. J. (2012). Rethinking Roman Britain: Coinage and Archaeology. Wetteren: Moneta.
Notes
1 See Gruber and Meadows 2014 for the original edition.
2 For recent compendia of evidence see Callataÿ, 1997a and 2003. For discussion of progress so far and the possibilities offered, see also eund., 2005 and 2011.
3 See Antike Fundmünzen Europa (http://afe.dainst.org/en/) for more information on German and Polish finds projects. Coin Hoards of the Roman Empire (http://chre.ashmus.ox.ac.uk). Coin Hoards of the Roman Republic (http://numismatics.org/chrr). Prototype for the Inventory of Greek Coin Hoards (http://coinhoards.org), which will see a major data and platform overhaul within the next two years as part of the NEH-funded Hellenistic Royal Coinages project.
4 Portable Antiquities of the Netherlands: https://www.portable-antiquities.nl/.
5 The National Endowment for the Humanities began funding Hellenistic Royal Coinages in 2017 (http://numismatics.org/neh-hrc2017), and the Joint Programming Initiative on Cultural Heritage announced its support for ARCH: Ancient Coinage as Related Cultural Heritage in December 2017 (https://www.greekcoinage.org/news--events/new-funding-for-online-greek-coinage-the-arch-project).
6 https://jena.apache.org/documentation/fuseki2.
7 http://nomisma.org/ontology.
8 SKOS (https://www.w3.org/2004/02/skos), WGS84 (https://www.w3.org/2003/01/geo), Dublin Core Terms (http://www.dublincore.org/documents/dcmi-terms).
9 NEH Grant PW-51711-14: https://securegrants.neh.gov/publicquery/main.aspx?f=1&gn=PW-51711-14.
10 Numishare, an open source framework hosted by Github and developed by Ethan Gruber: https://github.com/ewg118/numishare.
11 Project management was provided by Meadows and Rick Witschonke; data for Republican coins types were created and generously supplied by Ian Leins and Eleanor Ghey at the British Museum and have been published in Coinage of the Roman Republic Online (http://numismatics.org/crro).
12 For more information on the development evolution of OCRE from its initial release July 2012 to its present, LOD-aware architecture, see Gruber (2013).
13 See links on the top of http://numismatics.org/ocre/id/ric.1(2).aug.543A for further details of the model. Machine-readable data are available as URLs in RDF/XML, Turtle, and JSON-LD serializations, as well as through HTTP content negotiation.
14 FOAF: http://xmlns.com/foaf/spec. VoID: https://www.w3.org/TR/void.
15 It should be noted that although the Highcharts library is still in use in Numishare as of the writing of this chapter, Nomisma.org uses d3.js. Numishare will eventually migrate to this more open and better supported library, most likely before the conclusion of the Hellenistic Royal Coinages project (2020).
16 How to Contribute Data: http://nomisma.org/documentation/contribute.
17 See Berners-Lee’s seminal work on “Linked Data” for more information: https://www.w3.org/DesignIssues/LinkedData.html. The Clean URL pattern is best explained at https://en.wikipedia.org/wiki/Clean_URL.
18 The Presentation API is documented at http://iiif.io/api/presentation/2.0.
20 The Reconciliation API for OpenRefine is documented at https://github.com/OpenRefine/OpenRefine/wiki/Reconciliation-Service-Api.
21 Two that are particularly useful are authored by Miriam Posner on her blog (http://miriamposner.com/classes/dh101f17/tutorials-guides/data-manipulation/openrefine-resources) and Seth van Hooland, Ruben Verborgh, and Max De Wilde, published by The Programming Historian (https://programminghistorian.org/en/lessons/cleaning-data-with-openrefine).
22 See more technical details about these features at http://numishare.blogspot.com/2017/10/nomisma-launches-openrefine.html, and the Reconciliation section of the Nomisma API (http://nomisma.org/apis#reconciliation) page contains links to further documentation.
23 The American Numismatic Society collaborated with Prof. Jere Bacharach in the publication of the Egyptian National Library’s collection of Islamic coinage online in 2015: http://enl.numismatics.org.
24 For more technical details about software architecture that underlies these APIs, as well as further discussion of the strategies for reconciling the University of Graz collection, see http://numishare.blogspot.com/2017/11/numishare-supports-openrefine.html.