Thursday, July 13, 2023

Corpus Nummorum integrated into

Corpus Nummorum integrated into

At the European Coin Find Network / meeting several weeks ago in Sofia, Bulgaria, I learned that Corpus Nummorum had a SPARQL endpoint. However, it is an endpoint that supports only the query protocol without a UI wrapper to make it more publicly usable. Sitting in the conference auditorium, I used the Nomisma SPARQL endpoint to execute the SERVICE protocol in order to query the CN endpoint. This revealed a triplestore that contains both types and coins, almost entirely conformant to the data model (there are some variations to model bibliographic references or source collections). However, federated queries are relatively slow. Upon arriving back home, I spun up a simple SPARQL UI wrapper (which functions exactly like Nomisma), that I had originally built as an interface on top of the old defunct British Museum RDF export Matt Lincoln had squirreled away. This UI wrapper works on any SPARQL 1.1-compliant endpoint, whether it was running on my locally installed instance of Fuseki or any accessible endpoint on the web (as is the case with Corpus Nummorum).

Corpus Nummorum SPARQL results in a GUI

After exploring the type and coin data a bit further, I created two SPARQL queries that implement CONSTRUCT to generate RDF/XML exports, one for types and the other for coins. I created an XSLT identity transform to slightly modify the structure of the RDF I extracted from the endpoint so that it would validate in the Nomisma RDF import system. I also replaced the static image URLs from the Bibliothèque nationale de France and American Numismatic Society with IIIF service URIs.

Most of the coins have images, although a significant number of examples in CN are photographs of plaster casts rather than the original object. The plaster cast photographs are modeled slightly differently in the RDF than our standard model, and I did not modify my SPARQL query to acquire these image links in the data I uploaded into Nomisma. Nor, at present, have I incorporated the plaster cast photos in the model CN developed, as the SPARQL queries for displaying images in Nomisma would have to be updated (and possibly also some code).

As far as I can tell though, the die axes, weights, and diameters for the coins represented by plaster casts measure the original metallic object, and are therefore accurate numbers to use for statistical analysis.

Enhancing Research Context within

Much like the recent integration of IRIS types into the LOD ecosytem, the upload of more than 11,000 coin types and 28,000 specimens from Corpus Nummorum fills in a sizeable gap in the context pertaining to individual Nomisma concepts related to the coinage of Thrace and the surrounding regions of Moesia Inferior, as well as Troas and Mysia in Anatolia. Most broadly, the Nomisma ID for Thrace is augmented with a paginated list of over 10,000 coin types, many of which are illustrated by at least one specimen, ranging from Greek types from Corpus Nummorum to Roman Imperial types in OCRE.

Drilling down into narrower concepts, such as mints or authorities, reveals more refined visualizations. The Kingdom of Odrysian Thrace produced 66 types from the mid-5th to early 2nd century BC, although only about 20% are attributed to two certain mints.

The map and types of Odrysian Thrace  

Some Caveats

Aside from the aforementioned lack of photographs of plaster casts, the Corpus Nummorum dataset includes some duplication of both types and coins that we should be aware about. Since the CN typology runs the gamut of coinages produced in Thrace and nearby regions throughout the duration of antiquity, there is some overlap between these types and other corpora, including Hellenistic Royal Coinages and Roman Provincial Coinage Online. CN types don't like to these other URIs, or vice versa, so queries may duplicate statistics from the same object. Similarly, since the CN database records all specimens internally rather than linking to external collections, coins from Nomisma partners that are already linked to HRC types that overlap with CN have also been duplicated in the CN export. The long-term solution here is for Nomisma partners to catalog their own collections with CN type URIs and then weed these items out of the Corpus Nummorum export. The canonical URI for a coin should be published by the holding institution rather than independent research datasets.


Duplicative types minted in Aenus and coins between CN and Seleucid Coins Online

There are about 500 coins from the ANS that have been given CN object URIs that are now available through Nomisma, but actually relatively few of these are duplicative of the coins the ANS has exported to Nomisma through links to HRC type URIs. However, we should incorporate CN URIs for non-Royal coinages into our own database so that we can export them directly. The same should apply for collections in Berlin's IKMK network.


No comments:

Post a Comment