Friday 10 June 2011

CLAROS joins the Pelagios Project

On the 17th of May 2011, CLAROS went live with a launch event at Wolfson College, Oxford. The project aims to "bring the world of ancient art onto the semantic web", and comprises the CLAROS Explorer, an image search interface, and a site which exposes the data as RDF. Data are provided by partners, who model their collections using RDF and the Erlangen OWL-DL 1.0 implementation of the CIDOC Conceptual Reference Model.

Shortly after launching we were invited to join the Pelagios consortium, and we're delighted to be on board.

What we've done so far



We're in the process of co-referencing all the places referred to by our partners. For example this ewer was found within the Ashmolean's conceptualisation of Henan province, which is (eventually) contained within a 'centralised' concept of China.

This allows us to find fifty things from Greece, Buddha-related objects from Thailand, and lots of places we believe are within Greece.

We've created a dump of our data should anyone else want to play with it. At some point soon we'll change the names of the graphs in the triplestore to match these, so that the graphs are also dereferenceable. In the interests of trend-setting, we've also got a semantic pingback server running on the off-chance that there are similarly-minded people out there.

What next



We need to keep co-referencing our places. We've done a lot of the low-hanging fruit (those that are mentioned a lot and/or are easily found), but from here on in it's the law of diminishing returns. So far we're linking against Pleiades and Geonames where we can, and using the CRM's "falls within" property to create a hierarchy of containment.

We've also discovered a few instances of places in the source data we don't know how to model. For example, "between Kermanshah and Dezful"; look these places up on a map and you discover they're 352km apart by road, so taking a colinear equidistant point isn't very helpful. As well as 'betweenness', we can't do uncertainty (e.g. "Athens (?)"), or proximity (e.g. "near Athens").

I still need to get my head around OAC and how that relates to what we're doing. At the moment we're using skos:closeMatch and skos:exactMatch to link our places to their Geonames and Pleiades counterparts. I had thought about using property reification vocabulary (PRV) to create something to attach provenance information to, but it's possible that this could all be handled within OAC. Suggestions welcome (and probably much needed!).