Transcript hider2008
Towards a semantic web Philip Hider This talk The Semantic Web vision Scenarios Standards Semantic Web & RDA Web 1.0, 2.0, 3.0 Internet to WWW (Web 1.0) Web 1.0 allows people to navigate the Internet easily, through hyperlinks Web 2.0 allows people to collaborate more on the Web Web 3.0 allows computers to find and use the data contained in Web documents Web 3.0 = the Semantic Web vision The Semantic Web vision It will allow computers to make sense of the content of Web documents, so that they can find and use this data independently Basis of SW already developed, with standards such as XML and RDF Like Web 1.0, it represents a bottom-up, distributed approach How would it work? Computers would be able to identify and ‘understand’ particular data in a Web document according to the metadata associated with that data metadata could be inside our outside the document Computers (agents) would then be able to relate that data to other data in other documents (or the same document) according to specified schemas, ontologies and rules They could then independently integrate data and process information according to tasks set by their human users A Semantic Web scenario User asks ‘Trip Agent’ to purchase the ‘best’ deal for a trip to New Zealand with date range x, family members y, time of day z, etc. etc. ‘Trip agent’ searches the Web for flights and accommodation, and is able to look up databases and specify conditions according to what it ‘knows’ about user’s preferences Semantic Web scenario Agent is able to ‘understand’ the deals available on different websites by integrating data from different sources, e.g. looking up geographic information systems (how far from the sea, shops, etc.), weather forecasts, family members’ calendars, etc. an ultimately suggesting the optimal combination of flight, hotel, tours, etc. Another scenario User asks if the latest Stephen King book is available in a nearby library, can’t remember what it’s called ‘Library Agent’ searches the Web for nearby libraries with books by ‘Stephen King’, finds a few different Stephen Kings, confirms with user which Stephen King, then identifies the latest novel via the official Stephen King website, but chooses the secondnearest library (by car) which holds it because of availability/format/library opening hours, etc. What do SW agents need? Information about the data, i.e. metadata, in a machine-readable format Including a shared understanding of the structure of that metadata and its relationship to other knowledge structures (ontologies) Some clever programming Standards for the Semantic Web Resource Description Framework Universal Resource Identifiers XML Unicode Schemas (such as XML schemas) Ontologies written in e.g. OWL Rules written in RIF, etc. SPARQL Resource Description Framework W3C standard A model used to structure resource descriptions Can be used to structure data about any kind of resource could be a book, or a car, or a flight ticket, or an experiment, etc. Based on ‘triples’, i.e. Resource – Property – Value (Subject – Predicate – Object) Universal Resource Identifiers For example, URLs And ISBNs People don’t have them yet OCLC working on ‘work identifiers’ Properties and some values are referenced as part of particular schemas, ontologies, etc. eXtensible Markup Language (XML) Another W3C standard More flexible than HTML, XHTML Can be used to encode any data Data can be in the same Web document or another document Can be used to express RDF, i.e. RDF/XML RDF/XML basis for metadata structures such as schemas and ontologies Schemas Standardised structures of resource description that define property elements in a taxonomic way Mostly based on a particular domain, e.g. pertaining to bibliographic data, or geospatial data, or flight booking data, or used car data, etc. Schemas Two main groups of schemas – XML schemas and RDFS (RDF schemas) Superseding Document Type Definitions (DTDs) Specific well-known schemas include Dublin Core ONIX RSS Some metadata encoded in RDF/XML <rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:foaf="http://xmlns.com/foaf/0.1/" xmlns:dc="http://purl.org/dc/elements/1.1/"> <rdf:Description rdf:about="http://en.wikipedia.org/wiki/Tony_Benn"> <dc:title>Tony Benn</dc:title> <dc:publisher>Wikipedia</dc:publisher> <foaf:primaryTopic> <foaf:Person> <foaf:name>Tony Benn</foaf:name> </foaf:Person> </foaf:primaryTopic> </rdf:Description> </rdf:RDF> Some metadata encoded in RDF/XML <rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:foaf="http://xmlns.com/foaf/0.1/" xmlns:dc="http://purl.org/dc/elements/1.1/"> <rdf:Description rdf:about="http://en.wikipedia.org/wiki/Tony_Benn"> <dc:title>Tony Benn</dc:title> <dc:publisher>Wikipedia</dc:publisher> <foaf:primaryTopic> <foaf:Person> <foaf:name>Tony Benn</foaf:name> </foaf:Person> </foaf:primaryTopic> </rdf:Description> </rdf:RDF> Ontologies More sophisticated than schemas, formalising more complex relationships between elements Also usually domain-specific Use extra languages, such as OWL, on top of RDF/XML etc. Ontologies give more scope for agents to be ‘clever’ Dublin Core can be expressed as an ontology or a schema What about MARC? MARC files are rather flat and do not readily define relationships between elements But can be expressed as an XML schema, i.e. MARCXML MODS is a lite version of MARCXML Mappings between MARCXML and other schemas (e.g. DC) Mappings Lots of them! Between different schemas, ontologies, languages, etc. AKA crosswalks By UKOLN, LC, OCLC, etc. etc. The more standards and adaptations, the more crosswalks Value sets Resource – Property – Value Schemas and ontologies may point to particular value sets, e.g. Book A hasaSubjectcalled DCterms:LCSH Apples where Apples is a value in the set of values known as LCSH In other words, they may point to controlled vocabularies SKOS Simple Knowledge Organization Systems SW standard for expressing controlled vocabularies such as subject thesauri http://www.w3.org/2004/02/skos Might promote use of LCSH, etc. Semantic Web & cataloguing More sophisticated use of library catalogues if they can be understood by Semantic Web agents Library resources more likely to be used in conjunction with non-library web resources SW about agents using cataloguing, not replacing cataloguing Semantic Web & RDA RDA is therefore aligning itself with DC and RDF RDA elements mapped to DC, ONIX, etc. DCMI/RDA Task Group RDA-DC application profile http://dublincore.org/dcmirdataskgroup Prospects for SW Examples of Semantic Web developments: http://www.w3.org/2001/sw/sweo/public/UseCases A lot of standards now in place, technology not so much of an issue With RDA, bibliographic domain ripe for SW take-up Pre-SW library work Post-SW library work Thank you. [email protected]