Featured Dataset: Government Art Collection

Posted on 09/30/2011 by

0


GAC LogoToday, we’re featuring a dataset which Leigh pulled together for our Culture Hackday last week. The British Government has been collecting works of art for the past century or so, and has a total of around 13,500 which is looked after under the Collection. On their site, they give a bit of context for reasons behind the collection:

Works from the Collection are displayed in the offices and reception rooms of several hundred major British Government buildings in the United Kingdom and around the world. Thousands of people visit these buildings every year and therefore the works of art themselves play a vital role in helping to promote British art and artists.

data diagramThe Government Art Collection dataset, contains metadata about the collection. The developer documentation outlines how these resources are modelled in the data. It was created through a crawl of the site, and is summed up:

The dataset contains over 10,000 art works from more than 3,000 artists. The art works depict nearly 2,000 places and over a 1,000 different people. The majority of the people mentioned in the paintings (over 700) have been linked to their description in Dbpedia. Over time the dataset will be updated to include links to other datasets.

Let’s have a look at the vocabularies and classes used in the set, now. We can quickly get a feel for the shape of the information, and also see which classes are most represented in the set:

Class Resources
http://xmlns.com/foaf/0.1/Image 19168
http://data.kasabi.com/dataset/government-art-collection/schema/ArtWork 10506
http://data.kasabi.com/dataset/government-art-collection/schema/Artist 3144

The dataset formed the basis of the annotation bookmarklet created during the hackday last week: When viewing a piece of Government Collection artwork (like this one) or a painting from Artfinder, you can annotate it with the name of the people it portrays. When the bookmarklet is clicked, the hack places the annotation box around the area it guesses is most likely a person’s face, and makes some suggestions about who is in the work. Behind the scenes, the hack is both reading and writing data, storing the information from each annotation in Kasabi.

For people who like SPARQLing for data, Leigh has provided a few good queries for selecting particular aspects of the set, and from the new dataset homepage layout, you can see them listed under the “Query” tab among the Default APIs. This query, for example, will find all the works by a particular artist:

PREFIX dct: <http://purl.org/dc/terms/>
PREFIX foaf: <http://xmlns.com/foaf/0.1/>
PREFIX gac: <http://data.kasabi.com/dataset/government-art-collection/schema/>
SELECT ?uri ?name ?image ?page WHERE {

  ?uri a gac:ArtWork;
     foaf:maker ?artist;
     dct:title ?title;
     foaf:depiction ?image;
     foaf:page ?page.

  ?artist foaf:name "Tracey Emin".
}

In SPARQL, the item you’re looking to match (in this case, Tracey Emin) needs to match exactly to the label within the dataset. If you’re looking for a more general result, you can use the Search API for a looser match. I used the trusty Pytassium command line tool to search for an artist I liked, and found Lowry:

@prefix foaf: http://xmlns.com/foaf/0.1/ .
 @prefix ns1: http://data.kasabi.com/dataset/government-art-collection/schema/ .
 @prefix rdfs: http://www.w3.org/2000/01/rdf-schema# .
http://data.kasabi.com/dataset/government-art-collection/artists/103939 a http://data.kasabi.com/dataset/government-art-collection/schema/Artist;
 rdfs:label “Laurence Stephen Lowry”;
 ns1:number “103939”;
 foaf:made http://data.kasabi.com/dataset/government-art-collection/works/15030,
 http://data.kasabi.com/dataset/government-art-collection/works/19902,
 http://data.kasabi.com/dataset/government-art-collection/works/27336;
 foaf:name “Laurence Stephen Lowry” .

I can then match the correct foaf:name label (who knew he’s Larry Lowry?) and get back results using the same query structure.

So, there we have it. A set based on artworks seen across UK civic buildings which has already been used in a hack. This could be a particularly useful set for other apps, and could also be used to help link up information about prominant British artists, works, and those portrayed. (You can even find the top-ten portrayed “sitters” using a simple query!)

Posted in: Datasets