simple knowledge organization system (skos) in the context of semantic web deployment, library of...

The Simple Knowledge Organization System (SKOS)in the context of

Semantic Web Deployment

Alistair Mileshttp://purl.org/net/aliman

Library of CongressMay 2008

THE FUTURE OF THE WEB

http://purl.org/net/aliman 2

• Testimony of Sir Timothy Berners-LeeCSAIL Decentralized Information GroupMassachusetts Institute of Technology

• Before the United States House of RepresentativesCommittee on Energy and CommerceSubcommittee on Telecommunications and the Internet

• http://dig.csail.mit.edu/2007/03/01-ushouse-future-of-the-web.html


I. Foundations of the Web• “The success of the World Wide Web, itself built

on the open Internet, has depended on three critical factors:

1.unlimited links from any part of the Web to any other;

2.open technical standards as the basis for continued growth of innovation applications; and

3.separation of network layers, enabling independent innovation for network transport, routing and information applications.”


A. Universal linking: Anyone can connect to anyone...

• “In simple terms, the Web has grown because it's easy to write a Web page and easy to link to other pages.”

• “What makes it easy to create links ... is that there is no limit to the number of pages or number of links possible on the Web.”

• “Adding a Web page requires no coordination with any central authority, and has an extremely low, often zero, additional cost.”


• “Adding a page provides content, but adding a link provide the organization, structure and endorsement to information on the Web which turn the content as a whole into something of great value.”


• “The universality and flexibility of the Web's linking architecture has a unique capacity to break down boundaries of distance, language, and domains of knowledge.”

• “These traditional barriers fall away because the cost and complexity of a link is unaffected by most boundaries that divide other media.”


• “The Web's ability to allow people to forge links is why we refer to it as an abstract information space, rather than simply a network.”


II. Looking forward

• “First, the Web will get better and better at helping us to manage, integrate, and analyze data.”

• “Today, the Web is quite effective at helping us to publish and discover documents, but the individual information elements within those documents ... cannot be handled directly as data.”


• “Today you can see the data with your browser, but can't get other computer programs to manipulate or analyze it without going through a lot of manual effort yourself.”

• “As this problem is solved, we can expect that Web as a whole to look more like a large database or spreadsheet, rather than just a set of linked documents.”


A. Data Integration

• “Locked within all of this data is the key to knowledge about how to cure diseases, create business value, and govern our world more effectively.”

• “The good news is that a number of technical innovations...

• ... (RDF which is to data what HTML is to documents, and the Web Ontology Language (OWL) which allows us to express how data sources connect together) ...

• ... along with more openness in information sharing practices are moving the World Wide Web toward what we call the Semantic Web.”


• “Progress toward better data integration will happen through use of the key piece of technology that made the World Wide Web so successful: the link.”

• “The power of the Web today, including the ability to find the pages we're looking for, derives from the fact that documents are put on the Web in standard form, and then linked together.”


• “The Semantic Web will enable better data integration by allowing everyone who puts individual items of data on the Web to link them with other pieces of data using standard formats.”


DATA WEBS FOR E-SCIENCE


FlyWeb Project

• Fruit flies (Drosophila ...) • Model organism• Extensive body of genetic research• Much of that knowledge is in journal papers• Recognised value of research data• Establish public databases

– E.g. FlyBase– Centrally curated


Data Webs

• Link data resources• Ask questions that no single data resource can

answer

• What’s the easiest, cheapest, most scalable way to achieve this?

• Agile approach, add value incrementally, return value early and often...



Vertical Web Apps

Vertical Web Apps

Level 0 – Any Data Resources in the WebLevel 0 – Any Data Resources in the Web

Level 1 – SPARQL End-pointsLevel 1 – SPARQL End-points

Level 2 – SPARQL End-points (Schema Alignment)

Level 2 – SPARQL End-points (Schema Alignment)

Level 3 – SPARQL End-points(Integrated Data)

Level 3 – SPARQL End-points(Integrated Data)

Web 2 Mash-ups

Web 2 Mash-ups

SPARQL Mash-upsSPARQL

Mash-ups

SPARQL Mash-upsSPARQL

Mash-ups

??????

Data Web Layer Cake

Example Application

• [insert screenshot of mashup]


Future, self-publishing

• As publishing data on the Web becomes easier...

• ...more research groups will publish their own data...

• ...rich network of data resources...• ...challenging traditional view of scholarly life

cycle & value chain ... value grid...


SKOS


Potted History

• SKOS 2001 (pre-alpha)– Thesaurus Interchange Format (TIF), LIMBER Project

• SKOS 2003 (alpha)– Semantic Web Advanced Development for Europe (SWAD-

Europe)• SKOS 2005 (beta)

– W3C Semantic Web Best Practices and Deployment Working Group (SWBPD)

• SKOS 2008 (W3C Recommendation)– W3C Semantic Web Deployment Working Group (SWD)



http://www.w3.org/2007/Talks/1211-whit-tbl/

Layers in the Web

• http://www.w3.org/2007/Talks/1211-whit-tbl/#(23)

• Third layer is network (graph) of connections beyond documents...

• ... people, organisations, genes, proteins, concepts ...

• Represent these connections (data) in the (Semantic) Web


KOS e.g. LCSH

• Can be viewed as a network of interconnected concepts

• Represent LCSH as data in the Web– Make those concepts and their interconnections

part of the Web


Publishing KOS in the Web?

• Use RDF– Basic framework for data in the Web – resources,

literals, links... (“graphs” of data)


Publishing KOS in the Web?

• Use SKOS• Standard set of...

– Resource types (Classes)– Link types (Properties)

• ... For representing KOS as RDF data• (N.B. Because use URIS as names for classes

and properties, call this an RDF vocabulary)


SKOS Resource Types (Classes)

• skos:Concept– E.g. Baseball in art

• skos:ConceptScheme– E.g. LCSH itself


SKOS Link Types (Properties)

• For labeling concepts– skos:prefLabel, skos:altLabel, skos:hiddenLabel

• For documenting concepts– skos:note, skos:scopeNote, skos:definition,

skos:editorialNote...

• For linking concepts– skos:broader, skos:narrower, skos:related


• http://inkdroid.org/bzr/lcsh/docs/slides/


Publishing LCSH in the Web

• Project LCSH into RDF (i.e. create an RDF representation)

• Publish it in the Web as linked data– http://lcsh.info

• Ed Summers, Clay Redding, Dan Krech, Antoine Isaac


Scope of SKOS

• SKOS will be an all-encompassing standard for the lossless representation and exchange of all varieties of knowledge organisation system ... ?– No

• http://lists.w3.org/Archives/Public/public-swd-wg/2008Feb/0116.html -- Antoine Isaac


• “...the things that we aim at representing are very diverse: some classification schemes use ‘codes’ and refer to ‘classes’, thesauri have ‘terms’ and so on.”


• “Yet, it happens, looking at the way these things are used now and will be in the near future (with more and more links established between them), that (i) some standardisation has to take place, and that (ii) this standardisation can be actually grounded on some observed practical similarities (http://www.w3.org/TR/skos-ucr/)”

• “Our aim is not to replace the original objects in their initial context of use, but to allow to port them to a shared space, based on a simplified model, enabling wider re-use and better interoperability.”


Lessons from the Web

• Less is more ...– E.g. REST over SOAP

• SKOS should capture a small amount of common ground ... Just enough to enable KOS’s valuable concepts and connections to be deployed in the Web and be linked to/from

• N.B. SKOS is infinitely extensible!– Easy to mix & match– Easy to refine


THE VALUE OF LINKS


The value of links

• The Web showed, links between documents are really useful

• Google’s pagerank showed, structure of network means something (and is worth something!)

• Social networking Web sites showed, how much we value other kinds of links


Linked Metadata

• You’ve got LCSH in the Web, what next?• ... Linked metadata...?


• http://inkdroid.org/bzr/lcsh/docs/slides/


• [insert demo, show how links change topology of information space]


Value Proposition

• Links are paths to the discovery of information• Links can be exploited in useful (and

surprising) ways• Well-established KOS like LCSH can be hubs in

the network of linked metadata, bridging ...• (On the Semantic Web, LCSH should get very

high semantic pagerank!)


USING URIS


Why use URIs?

1. Identifier management2. Data discovery


Identifier management

• Referring to things• In a database, each table has a primary key• What happens when you try to combine data

from 2 databases?– Identifier clashes (ambiguous reference)– Identifier aliases (co-reference)

• Clashes hurt precision, give you nonsense• Aliases hurt recall, miss important results/links


URIs & identifier management

• URIs are like a single, global pool of identifiers – one world-wide primary key

• Can claim ownership of parts of “URI space”• Even though we’re all using same primary key,

mechanism for avoiding URI clashes• Can join data from multiple sources with

confidence• But ... doesn’t solve the alias problem, still

need to find co-references


Data discovery

• Problem with distributed data ... How do you find everything thats “out there”?

• Two general approaches:– Centralised– Decentralised


Centralised discovery

• Someone somewhere keeps a “catalogue” of everything

• Everyone “knows” where that catalogue is• New sources “tell” the catalogue about

themselves (a.k.a. “register” themselves)• E.g. Gas maintenance• Works well at small-medium scales

• E.g. FlyWeb• Rely on networks outside the Web (e.g.

Knowing the right people)http://purl.org/net/aliman 57

FlyWeb Project

• [small number of large data resources]


Decentralised discovery

• Data in one source refers data in another (using a URI)

• Data from the other source can be retrieved directly, by “de-referencing” the URI via the Web

• So given one data source, you can “follow your nose” and retrieve data from all linked sources ...

• ... without needing a central catalogue or registry, just the Web

• Works well up to Web-scale– E.g. FOAF


Dereferenceable?

• For some URIs, can retrieve a “representation” of the “resource” via the Web

• (N.B. “resource” = “thing”)


FOAF

• Very large number of relatively small data resources


Why use URIs?

• Identity management– From 2 to 2 billion data sources, always a problem

• Data discovery– Ability to “de-reference” a URI opens possibility

for decentralisation– Ability to “de-reference” is also useful in

centralised models (e.g. Registries can harvest)


SEMANTIC WEB DEPLOYMENT


W3C SWD WG

• SKOS• RDFa• Recipes for publishing RDF (linked data)• Vocabulary management


W3C Semantic Web Activity

• Semantic Web Deployment• Data Access (DAWG)

– SPARQL query language, SPARQL protocol

• GRDDL• OWL 2• SWHCLSIG• SWEO


SUMMARY


Suggestions

• Linked KOS– Project LCSH into RDF (SKOS) – done – Publish LCSH as linked data in the Web – done– Publish SPARQL endpoint for LCSH

• Linked metadata– Project LOC metadata into RDF– Publish LOC metadata as linked data in the Web

• With links to LCSH & LCC– Publish SPARQL endpoint for LOC metadata


Bibliographic Information as RDF?

• Projecting LCSH into RDF ... SKOS is standard vocabulary of resource & link types

• Projecting LCSH metadata into RDF ... Which vocabulary to use???

• Challenge – diversity of bibliographic information!


RDA -> RDF

• Joint DCMI/RDA task force• Seed funding to develop initial prototype RDF

vocabularies for bibliographic information• Based on FRBR and data model implicit in RDA• Early stages• http://dublincore.org/dcmirdataskgroup/• Karen Coyle


Thanks

• STFC Rutherford Appleton Lab– Brian Matthews, Michael Wilson, Juan Bicarregui

• Oxford Image Bioinformatics Research Group– David Shotton, Graham Klyne, Jun Zhao

• W3C Semantic Web Deployment WG• Members of [email protected]

• Comments on SKOS -> [email protected]://purl.org/net/aliman 70