New Search | Return to Search Results | Print View

Poster: CLAROS—Collaborating on Delivering the Future of the Past

Rahtz, Sebastian, University of Oxford Computing Services,

Dutton, Alexander, University of Oxford Computing Services,

Kurtz, Donna, Beazley Archive, University of Oxford,

Klyne, Graham, Department of Zoology, University of Oxford,

Zisserman, Andrew, Department of Engineering Science, University of Oxford,

Arandjelović, Relja, Department of Engineering Science, University of Oxford,

CLAROS (Classical Art Research Online Services) is a technology and data collaboration between classical art and archaeology research projects, museums and semantic web researchers. Documenting objects from the museums of the world, CLAROS aims to engage with the public across the widest possible spectrum. It builds on the success of the Beazley Archive which has provided programmes for the public as well as the scholar and an illustrated linked dictionary for more than fifteen years. CLAROS ( is based at the University of Oxford, with partners in the UK, France, Germany, Switzerland and Greece. The portfolio includes an aggregating RDF database, web discovery interfaces for different types of audience, visual search using image analysis of shapes and images, semantic information extraction from digitized text, place and name gazetteers, and investigation of avatars for resource discovery. CLAROS aims to be an effective and powerful partner in the realm of semantic web and linked data about the past.

Data Modelling

The CLAROS aggregating data cache pulls information from its partners, limiting itself to those interchange components which can be mapped to the CIDOC Conceptual Reference Model (CRM) ontology, with a few extensions relating to dates and geolocations. The majority of records so far use CRM concepts E22_Man-Made_Object, E53_Place, E52_Time-Span and E21_Person. The data contributions, using XML RDF as the ingest form, include the Beazley Archive (pottery and gems) and the Lexicon of Greek Personal Names (onomastic data) at the University of Oxford, the Arachne Sculpture Archive, the Lexicon Iconographicum Mythologiae Classicae in Paris and Basel, and the German Archaeological Institute, producing an RDF triple store of over 20 million assertions. The federating and subsidiarity principle of CLAROS is that it acts simply as a resource discovery system, with search results linking back to the host database for more information, preserving the IPR and intellectual integrity of each partner. A SPARQL interface and RESTful APIs are provided for expert use, but CLAROS itself provides an exemplar query and visualisation interface (the Explorer) with an emphasis on textual search, timeline display and mapping of results. It is expected that the user will start with broad search terms, receive back information from a wide variety of sources, and then gradually refine and explore the results, perhaps ending in unexpected places.

Full Size Image


One of the research subprojects of CLAROS is Metamorphoses, whose aim is to establish a working co-reference system for name, place and date information in classical art and archaeology and ancient history projects at Oxford. This will provide geo-naming and geo-locating services to both CLAROS partners in their normal research, and to the CLAROS Explorer in performing searches. One output is an aggregating Ancient Place Server with web services to answer queries about locations of places, names of places, and types of places within a chosen area. We work on a social model of places as objects which come into existence as a result of naming by one or more groups of people at a particular time. We expect places to have multiple names, to have different geographical limits over time, and to have relationships with other places.

Visual Search

A novel aspect of CLAROS is that visual queries can be used to access and search the archives that are linked to within CLAROS. Suppose, for example, that a novice takes a photograph of a Greek vase in a museum on their iPhone, or finds an image of a Greek vase on the web, that they would like to know more about. This image can be use as a visual query to retrieve that vase from the archive, in much the same way as the text phrase “Greek vase” can be used as a query in Google to retrieve web pages which contain that phrase. The method is illustrated in Figure 1: the image is uploaded to a web server, and the server returns the matches in the archive together with meta-information – for instance identifying the type of the vase, its date, its material, its decorations etc.

Full Size Image

Full Size Image

Full Size Image

Figure 1: Identifying a vase from a web photo: an image of a vase is uploaded from a web link (top), and is matched to one in the Beazley vase collection (middle) in real time (a search through over 100,000 images). This identifies the type of vase, its material, date and decorations from the meta-information (bottom) associated with this data in the Beazley archive.

Figure 2 gives another example, this time for a visual search of the Arachne classical sculpture archive.

Full Size Image

Full Size Image

Figure 2: Identifying a sculpture from a web photo: an image of a sculpture is uploaded from a web link (top), and is matched to one in the Arachne sculpture collection. In turn, this provides links to other images of the sculpture in the collection and also to associated meta-information (bottom).

That this visual search is possible, and indeed can be carried out with results being returned immediately, is due to recent methods developed in the computer vision community on visually searching for objects in large scale image datasets (see [1] for details).


J. Philbin, O. Chum, M. Isard, J. Sivic, and A. Zisserman 2007 “Object Retrieval with Large Vocabularies and Fast Spatial Matching, ” IEEE Conference on Computer Vision and Pattern Recognition,