Has been described as a ‘data commons’, or more usually a Web of Data.
Linked Data and the OpenART Project Tate, 7th October 2011 Julie Allinson, University of York Richard Stephens Steve Baylissand Martin Dow, Acuity Unlimited
“The term Linked Data refers to a set of best practices for publishing and connecting structured data on the Web.” “the Semantic Web is the goal or end result… Linked Data provides the means to reach that goal” From ‘Linked Data: The Story So Far’ - Heath, Bizer and Berners-Lee 2009
The goal of Linked Data is to enable people to share structured data on the Web as easily as they can share documents today. Bizer/Cyganiak/Heath Linked Data Tutorial, linkeddata.org
Haven’t we been putting linked data on the web for years?
The web is full of data, yes, but structure in traditional web pages is largely ‘formatting’
Databases, search interfaces, XML etc are not really integrated into the web, just on it
Web 2.0 services tend to work against a fixed set of sources
links to galleries links to exhibitions
Linked data Linked data is about identifying the ‘real-world’ things we want to describe and exposing rich information about these ‘things’ in the form of links linked data applications operate on top of this unbound structured web of information the www as one big database
RDF RDF (Resource Description Framework) is a way of modelling data, conceptually (as graphs) it links things, not just documents and creates ‘typed’ links allows us to say meaningful things about entities and the relationships between them
Triples Triples are the basis of RDF Triples statements Subject – Predicate - Object Rather than Displays Painted Links to The Snail The Snail Gallery Matisse Web Page Web Page
One more thing about triples … If we want to do ‘linked’ data. Each piece of the triple should be a link. In web speak, that means a HTTP URI
For discussion? Quality and Trust Licensing Control and Loss of Control Hype Curve Other approaches
OpenART : Open metadata for Art Research at the Tate JISC funded part of their Infrastructure for Resource Discovery Strand 6 months (March to July) – short! £30k – cheap! Building on the (much bigger) York’s History of Art Department and Tate’s Court Country City project funded by the AHRC. The Digital Library are leading OpenART and providing technical development to Court Country City Acuity Unlimited are working with us on OpenART.
Doing what? designing and exposing linked open [meta]data for an important research dataset entitled "The London Art World 1660-1735", created as part of the AHRC funded Court, Country, City: British Art 1660 – 1735 project
And again, doing what? ‘The London Art World 1660-1735’ is set of spreadsheets, the result of several years research into Sources (sale catalogues, newspaper adverts) of information about Sales, People, and Places (and the things being sold)from the period The Digital Library is already helping put this information on the web. OpenART is extending this to create ‘linked open data’.
OpenART - in brief Take the spreadsheets Anaylse them and model into an ontology Decide how to create the linked data and how to expose it to the web Experiment a bit Test and validate the ontology Create the data and expose
Describing things with ontologies an ontology is a standardized representation of knowledge as a set of concepts within a domain, and the relationships between those concepts. It can be used to reason about the entities within that domain, and may be used to describe the domain (wikipedia) or, put another way “a way of describing things in a way other people/machines can understand” (me)
There are many ontologies out there Dublin Core is a simple ontology for describing web resources Geonames describes places For OpenART we have used two core event-based ontologies Dolce Ultra-lite Linked events (LODE) and have extended these for the domain
Example! Much of the OpenART data set is about art sales, which are ‘events’ Linked Events Ontology has a class for ‘Event’ Which we have specialized to create an OpenART class for ‘Sale’ DUL has a class for ‘Social Agent’ (or person) Which is a superclass for our term ‘Organiser’
Lost yet? What this means is that we can describe our dataset for a very narrow domain (people researching the art market) But it can be generalised and understood much more widely Because our narrow classes are members of broader, more general classes used in standard ontologies
Example (again)! Linked Events Ontology has a class for ‘Event’ Which we have specialized to create an OpenART class for ‘Sale’ Someone only looking for ‘art sales’ in 1670 can answer their research question Someone looking for ‘events’ in 1670 could infer from our ‘Sale’ class that these sales are also events
OpenART output Unique identifiers for the primary entities (Sale, Source, Person, Place, and Artwork) RDF/XML (and probably ttl) documents representing each primary entity RDF/A embedded in the CCC web site Owl:sameas statements for personal names and VIAF
Lessons learnt Ontology development is hard work and needs sustaining and a community Linked data is simple in principle, but in application needs expertise and commitment The original data is semi-structured and this needs clean-up to fit the precision of the ontology
Final thoughts I believe there is a lot that could be done with open and linked data, by institutions … not least efficiencies, adding-value to local information, information sharing and resource discovery … Contact Julie Allinson email@example.com http://tinyurl.com/dlib-openart -- http://yorkdl.wordpress.com/ (thanks to Adrian Stevenson for some of the slides; cover image: http://www.flickr.com/photos/bostonjal/5280851979/)