Semantic Web
From MediaFranca
Contents |
Introduction
The Semantic Web is slowly emerging as a new standard for sharing knowledge and resources. The benefits of this new technology are enormous, especially significant are the easy way of sharing knowledge and the ability of merging graph representations in the form of metadata and ontologies. The problem though is that its architecture is too complex for people to understand and collaborate. There seems to be a structural lack of concern with the end users, which has signified that this new kind of standard is not likely to success in the short term. I propose the development of a comprehensive visual interface to serve as a pivot between end users and the network in order to bring value to the Semantic Web.
What is it?
Semantics: signifying meaning.
The Semantic Web is about making webpages readable by computers. This would imply that computer could look at severeal webpages and then combine that information modelling fairly complicated data structures. This capacity of merging different data structures allow the computer to inferencing or reasoning, in this way computers can turn into semi-intelligent agents by parsing webpages.
Ontology: Specification of a conceptualization. Basically is a vocabulary for classes and set of relationships. The simplest ontology language is RDFS (RDF schema)
RDF Schema
- Classes
- rdfs:Resource
- rdfs:Class
- rdfs:Literal
- rdfs:Datatype
- rdf:XMLLiteral
- rdf:Property
- Properties
- rdfs:range
- rdfs:domain
- rdf:type
- rdfs:subClassOf
- rdfs:subPropertyOf
- rdfs:label
- rdfs:comment
Notes
Knowledge Representation and (vs) the Semantic Web by Peter Patel-Schneider. January 25, 2006
- Current databases are Formal Symbolic Representations (FSR)
- Google does information retrieval, and it's not FSR
- The semantic web is a good place to make FSR (or isn't?)
- The SW is an extension of the current web where information is given well defined meaning
- The semantic web is like a robot were you don't need to worry about vision and motion, and his sensors are represented through it's services
- IRI instead of URI (international is more universal that "universal" in the web)
- The semantic web is about representation logics (logic frameworks)
- part of it exist nowdays in limited amount of things
- RDF is the underlying language for the SM, distinguished from XML which represents data tied to docs from transmitter to reciever. RDF is design to represent info that anybody can use.
- RDF is transported as XML, however
- RDFS extends this with a tiny ontology language (like a special meaning, specialization versus generalization)
- An ontology is a way of roughly describing the categories that are in the world
- Why building languages in the semantic web is broken?
- SW are several languages (XML, RDF, OWL) that fit in a particular way (initial version, late 1990s)
- Later development (2001-2004)
- ideas as objects, i.e love
- properties turned into relations
- formal definitions of meaning (classes and subclasses)
- missing definitions for names, i.e. John
- Single parser for the SW - parses RDF/XML into RDF triples
- this things breaks down when we try more expresive languages
OWL
- overcomes these kind of problems and has well defined meaning
- Models theoretics semantics and has relative relationships
- makes inferences and works with conjunctions
- you can build reasoners from this language
Three problems:
- triples, triples and triples
- they are the only syntax
- triples are always facts
- autereferential paradoxes (the data is part of itself, the Russell paradox)
- triples don't distinguish between users and names (a conflcit between classes and objects)
How can we make the SW more useful?
- Push for a unified vision at the W3C (politics first)
- Select and design services language
- It's hard to author ontologies
- lack of human authoring of ontologies
- it's hard to write formal stuff
- tough HCI problem
- could we build semi-automatic tools for extracting knowledge?
- lack of human authoring of ontologies
