Embed Size (px)
Citation preview
+
RDFSA little semantics goes a long way
Mariano Rodriguez-Muro, Free University of Bozen-Bolzano
+Disclaimer
License This work is licensed under a
Creative Commons Attribution-Share Alike 3.0 License (http://creativecommons.org/licenses/by-sa/3.0/)
A few examples from these slides has been taken from Semantic Web for the working Ontologist. Chapter 6.
Some of the slides on the use of taxonomies are based on: http://info.earley.com/webinar-replay-business-value-
taxonomy-aug-2012
+Reading material
Semantic Web for the working Ontologist. Chapter 6http://proquest.safaribooksonline.com/book/-/9780123859655
+
Overview
Motivation
Ontologies
RDFS ontologies
Annotations
+
Motivation
+Motivation
Motivations for semantic technology Making the web machine
understandable Expressing knowledge Reasoning with
knowledge
However, points 2 and 3 are not possible with the technologies seen so far
RDF doesn’t define vocabularies, and
Different datasets may use different URI’s to represent the same kind of data
+Motivation
Agreement in RDF graphs concerns only the data model and the use of URI as identifiers
No semantics!
+Ontologies and Ontology Languages
+What is an ontology language?
Specification of valid “axioms”
Specifications of vocabularies with “predefined” meaning in axioms
Informal: Topic Maps, UML diagrams
Formal examples: Predicate Logic, First Order Logic
Semantic Web examples: RDFS SWRL, OWL
Different languages have different expressive power
Axioms allow to produce “inferences”
The more expressive power, more complex and costly the inferences
+What is an ontology?
Collections of “axioms”
Describe the meaning of the vocabulary of a domain (e.g., an area of expertise)
Expressed in an Ontology Language
Valuable on their own as knowledge repositories
In combination with data valuable to implement complex behavior with little or no coding
+Example: Schema.org
Schema.org IS a simple ontology
Organizes terms in hierarchies with predefined meaning
The language is a variation of RDFS
+
RDFSIntroduction by example
+RDFS
W3C standard for an ontology language
RDFS introduces resources (URIs) with a predefined meaning
Inference engines that support RDFS allow to take that meaning into account
RDFS inferences extend the RDF graph by means of inference and hence, affect query answering
RDFS is very simple compared to SWRL or OWL, however, it is very useful in many context, allowing for increased productivity, easy data integration and interesting AI applications
+Building blocks
New namespace rdfs: <http://www.w3.org/2000/01/rdf-schema#>
New categories: Classes, resources that share something in common, allow
us to group things together. For example, Employee, Company. Resources that identify classes have rdf:type rdfs:Class
Instances, resources that are “members” of a class
Commonly, Class names are nouns
+Building blocks
Resources can belong to multiple classes
+Building blocks (cont.)
Properties: Resources used as a predicate in statements
Commonly, Property names are multiple words, expressing direction and in camel-casing
+RDFS Ontologies
RDFS Axioms Are RDF triples!
RDFS ontology is an RDF graph!
An RDF graph may have a subgraph expressed in RDFS We call the RDFS axioms/triples the Tbox of the ontology
(terminological information, predefined meaning) The rest is the Abox of the ontology (plain data, no
predefined meaning)
+Type propagation
RDFS vocabulary:rdfs:subClassOf
Key notions sub class (on the left) super class (on the
right)
Intuitive meaning, if :mariano is an instance of subclass it is also an instance of superclass
Formal meaning: subsets
Inference: type propagation
Similar to inheritance in Object Oriented formalisms
+Type propagation
RDFS vocabulary:rdfs:subClassOf
Key notions sub class (on the left) super class (on the
right)
Intuitive meaning, if :mariano is an instance of subclass it is also an instance of superclass
Formal meaning: subsets
Inference: type propagation
Similar to inheritance in Object Oriented formalisms
+Relation propagation
RDFS vocabulary:rdfs:subPropertyOf
Key notions sub property(on the left) super property(on the right)
Intuitive meaning, if (x,y) are connected with subproperty they are also connected with superproperty
Formal meaning: subsets (of binary tuples)
Inference: relationship propagation
+Relation propagation
RDFS vocabulary:rdfs:subPropertyOf
Key notions sub property(on the left) super property(on the right)
Intuitive meaning, if (x,y) are connected with subproperty they are also connected with superproperty
Formal meaning: subsets (of binary tuples)
Inference: relationship propagation
+Types by usage
RDFS vocabulary:rdfs:domain, rdfs:range
Key notions domain of a triple:
the subject range of a triple:
the object
:p rdfs:domain :C > the domain of any triple where :p is the predicate is an instance of :C (similar for rdfs:range)
Formal meaning: if (x,y) in P, then x in :C
Inference: type assignment by property usage
+Types by usage
RDFS vocabulary:rdfs:domain, rdfs:range
Key notions domain of a triple:
the subject range of a triple:
the object
:p rdfs:domain :C > the domain of any triple where :p is the predicate is an instance of :C (similar for rdfs:range)
Formal meaning: if (x,y) in P, then x in :C
Inference: type assignment by property usage
+Interactions
All inferences interact to allow complex behavior
+Interactions
All inferences interact to allow complex behavior
+Set intersection
Proper set intersection is not possible in RDFS
However, expressing necessary membership to multiple classes is possible, i.e., A subset B AND C
A rdfs:subClassOf BA rdfs:subClassOf C
considerx rdf:type A
+Set intersection
Proper set intersection is not possible in RDFS
However, expressing necessary membership to multiple classes is possible, i.e., A subset B AND C
A rdfs:subClassOf BA rdfs:subClassOf C
considerx rdf:type A
+Set intersection
Proper set intersection is not possible in RDFS
However, expressing necessary membership to multiple classes is possible, i.e., A subset B AND C
A rdfs:subClassOf BA rdfs:subClassOf C
considerx rdf:type A
One direction only!
+Set intersection
Similar for roles
+Set intersection
Similar for roles
+Set union
Proper set union is not possible in RDFS
However, A OR B subsetOf C
B rdfs:subClassOf AC rdfs:subClassOf A
considerx rdf:type B
orx rdf:type C
+Set union
Proper set union is not possible in RDFS
However, A OR B subsetOf C
B rdfs:subClassOf AC rdfs:subClassOf A
considerx rdf:type B
orx rdf:type C
+Set union
For roles. Aligning to a global vocabulary
+Set union
For roles. Aligning to a global vocabulary
+Equivalence
Merging vocabularies
To account for same use of different terms (classes or properties)
For classes or proeperties
+Equivalence
Merging vocabularies
To account for same use of different terms (classes or properties)
For classes or proeperties
+Last notes on RDFS axioms
Main new vocabulary: rdfs:subClassOf rdfs:subPropertyOf rdfs:domain rdfs:range
Different from CONSTRAINTS, missing triples are NOT a violation
Allow to infer new information
Allows to implement system behavior!
+Open lists revisited
RDFS also facilitates access to Lists
Elements of lists are a possibly infinite set of elements of the form rdf:_1, rdf:_2, etc
Access difficult in practice
RDFS facilitates this by enforcing that: if x rdfs:_1 ythen x rdfs:member b
+Open lists revisited
RDFS also facilitates access to Lists
Elements of lists are a possibly infinite set of elements of the form rdf:_1, rdf:_2, etc
Access difficult in practice
RDFS facilitates this by enforcing that: if x rdfs:_1 ythen x rdfs:member b
More detail on this on the lecture aboutRDFS semantics
+Axiomatic triples
RDFS enforces certain facts to be always true
These facts are statements (triples)
Referred as Axiomatic triples
Listed in http://www.w3.org/TR/rdf-mt/
More detail on this on the lecture aboutRDFS semantics
+RDFS Semantic Conditions
Every resource xx rdf:type rdfs:Resource
Every literal xx rdf:type rdfs:Literal
… etcMore detail on this on the lecture aboutRDFS semantics
+Last notes on RDFS axioms
Main new vocabulary (not the only one): rdfs:subClassOf rdfs:subPropertyOf rdfs:domain rdfs:range
Different from CONSTRAINTS, missing triples are NOT a violation
Allow to infer new information
Allows to implement system behavior!
+
Hands on examplesFrom Semantic Web for the Working Ontologist
+Automatic classification of employees (part 1) Transform into an RDF representation
Automatically catalog objects as Employees, and as Active employees, Suspended employees and Ex-employees using a minimal set of “axioms”
<ID> Project Assignment
AbsentUntil
Termination Date
22 24 - -
34 24 Dec 23, 2012 -
73 - - Jun 4, 2010
Employee table. Primary key: 10Active employees are assigned to projects
+Automatic classification of employees (part 2) Transform into an RDF representation
Automatically catalog objects as Employees as managers
<ID> Project Name
<Manager>
24 Project-x 34
25 Project Mayhem
22
Project table. Primary key: IDForeign key <Manager> to Employee table
+Align vocabularies
• Align corresponding properties using RDFS
• Align with FOAF vocabulary (when possible) using RDFS (use foaf:name, foaf:homepage)
+
Annotations
+Annotations
URI’s are not readable
Readable information (comments, names, etc.) can be stored using properties, but
Property names are not standard, however, we could like some standard names for “human oriented information”
RDFS defines: rdfs:label
A readable name for a resource rdfs:comment
Human focused comments
These are propertiesSo, subPropertyOf can be used with them
+Redirection
Redirecting to location of documents (RDF) with additional information about a subject
No formal semantics
RDFS provides: rdfs:seeAlso. Additional information rdfs:definedBy. Authority information, primary source.
These are propertiesSo, subPropertyOf can be used with them
Recall the semantic webidea, linked databases
