Transcript
OWLRepresenting Information Usingthe Web Ontology Language
1 Current Web
•Publishing medium•Dominated by HTML
▫Hyper Text Markup Language•Pages accessible using URLs
▫Uniform Resource Locators▫http://www.w3.org/
•Supports human readers using browsers
1.1 Current Web History
•Internet infrastructure created by DARPA•Mostly text-based (telnet, ftp, gopher)•1992: Tim Berners-Lee/CERT developed
▫HTML & HTTP (Hyper Text Transfer Protocol)
▫Web browser (Mosaic)•Allows anyone to publish structured
documents connected by hyperlinks•Combined with TCP/IP and XML
(eXtensible Markup Language) to create “killer app”
1.3 The Web is Not Enough
•Not enough structure to support computer processing of content
•No way to connect information to enable complex queries
•HTML too focused on format/display•Need to add markup to explain meaning
(semantics) •Semantics will enable automated
interpretation of structured web content
1.3.1 Information Structure
•HTML documents▫Semi-structured formatting▫Unstructured text
•Natural Language Processing (NLP)▫Improving, but impractical on a large scale
•Structured database information must be shared in a computer-parseable maner
•Goal: allow automated software agents to mine the web, creating new functionality
1.3.2 Finding Requires Metadata “Find the cheapest Key lime pie within 5 miles.”
• Keyword-based search engines▫ Find pages that might contain desired content▫ Don’t provide answers to questions…the goal!▫ Have to find local restaurants, then look at their
menus• Query engines aim to answer questions
▫ Should be able to filter restaurants within 5 miles, access menus, compare prices, get answer
▫ Show how answer gotten from reliable sources
1.3.3 Semantics Must Be Explicit
•Providing semantic information explicitly in documents enables software to:▫Manipulate information (filter, summarize)▫Infer new facts (inference)▫Link multiple distributed information
representations (semantic join)
2.2.2.1 Structured Representations
•Computers need▫Consistently structured information
collections▫Inference rules to conduct automated
reasoning▫Representations formal enough to detect
inconsistencies and errors▫Network-distributed information to support
scalability
2.2.2.2 Supporting Language
•Need a tagged markup language to provide▫Syntax
Language format rules; open & vendor-neutral
▫Semantics Meaning of concepts; formal, finite, &
extensible▫Expressiveness
Richness; able to express concepts & relationships
Completeness, correctness, & efficiency (hardest!)
▫Standards Common language for all
2.2.3 Compromise
•Must balance need for structure with need for human-friendly data representations▫True natural language processing not yet
ready▫Humans don’t like to process raw
structured data•Proposed solution
▫Humans must augment content with markup
▫Must show an ROI payoff for extra effort
2.3 Semantic Web to the Rescue•Next evolutionary generation of the web
▫Structured information representations provide explicit meaning
▫Information “marked up” according to language standards
▫Software provides new functionality by interpreting, exchanging, & processing meaning
•Technologies focus on information representations tied to explicit meaning
2.3.1 Semantic Web History•Term coined by Sir Tim Berners-Lee•US Dept of Defense/DARPA created
DAML▫DARPA Agent Markup Language▫Helped define critical concepts
•European Union created OIL▫Ontology Interface Layer▫Combined with DAML to create DAML+OIL
•W3C built on DAML+OIL to create OWL▫Web Ontology Language (yes, it’s out of
order)▫First draft approved February 2004
2.3.2 Semantic Web Vision
•Next generation of the web•Vast object-oriented integrated
knowledge base that can be accessed and inferenced via machine-understandable schemas
•Transparent to the end-user•Link documents and the information in
them•Leverage the current web infrastructure•Reduce the cost of performing tasks
2.3.4 Use Cases
•Tactical level functionality▫Lower-level functions & basic operations▫Behind the scenes
•Strategic applications▫Higher-level compositions of tactical
features▫Provide more complex functionality▫Customer-facing
2.3.4.1 Tactical Services
•Describe distributed information▫Harvest content, process, & exchange
results•Support queries
▫Answer questions & explain reasoning•Support searching
▫Find information based on meaning, not keywords
•Support inferring▫Drawing conclusions from explicit facts▫Reduces size & complexity of knowledge
bases
2.3.4.2 Strategic Applications•Vertical applications
▫Provide specialized services to a particular domain
▫E-commerce (B2B, B2C)•Agent software
▫Autonomous; mobile; architecture-independent
▫Find & interpret information, act, report results
•Information management▫Migrate intelligence from the software to
the data▫Provide new functionality without
modifying code▫Integrate repositories
2.3.5 Appropriate Applications
•Semantic web applications appropriate to:▫Publish content for both humans and
computers▫Share information without understanding
model▫Inferring new facts & joining information
sources•Characteristics of good candidate
domains:▫Well-understood but dynamic domain▫Heterogeneous information sources▫Existing information interchange
requirements•Not suited to binary data, e.g. image
processing
Chapter 3
3.1 Ontology Definitions•Historical definition
▫Studies of the science of being, and the nature and organization of reality
▫Definitive classifications of objects & their relationships
•Other definitions▫Computer science definition▫Types of ontologies▫Gruber definition▫OWL-specific ontology definitions
3.1.1 Computer Science Definition• Popularized by AI
community • Tbox
▫ Terminogical components▫ Equivalent to “schema”▫ Define concepts▫ Semantic Web equivalent
Ontology• Abox
▫ Assertional components▫ Equivalent to “records”▫ Assert facts▫ Semantic Web equivalent
Individuals
3.1.2 Types of Ontologies• Many types
▫Domain ontologies▫Metadata ontologies (Dublin Core)▫Method/task ontologies
• Many ways to classify ontologies▫Formality▫Regularity▫Expressiveness
• Simplest ontology: Taxonomy▫Hierarchy of concepts related with IS-A
relationship▫Can’t express complex relationships
3.1.3 Gruber Definition
• “Formal specification of a conceptualization” – T. Gruber
• An ontology is a▫ Formally-described▫ Machine-readable▫ Collection of terms &
their relationships▫ Expressed in a language▫ Stored in a file
3.1.4 OWL-Specific Ontology Def’n•Web Ontology Language (OWL) ontology
▫“An OWL-encoded, web-distributed vocabulary of declarative formalisms describing a model of a domain”
•Domain▫A specific subject area or area of
knowledge▫Typically the focus of a particular
community of interest•Encode a model of the domain, not all of it
3.2 Ontology Features
•Communicate a common understanding of a domain• Example: restaurant association describes relationships between food
items
•Declare explicit semantics• Make assumptions explicit• Reduce ambiguity
•Make expressive statements• Have reasoning properties to support scalable, decidable inferencing
•Support sharing of information• Allow semantic mapping between information sources
3.3 Ontology Development Issues• Authoring ontologies
▫Can be developed by anyone, but▫Better if developed by consensus-based
standards development groups▫Vertical ontologies describe a domain▫Horizontal ontologies span domains and
describe basic concepts• Separating ontologies from individuals
▫Usually a good idea▫Sometimes not possible
• Committing to an ontology▫Makes applications easier to understand,
modify, reuse
3.4 Describing Semantics
•Defining information representation building blocks
•Describing relationships between building blocks
•Describing relationships within building blocks
3.4.1 Building Blocks• Three basic blocks
▫ Class constructs▫ Property constructs▫ Individual constructs
• Together, they describe a model of a domain
• Each type requires▫ A computer-
understandable representation
▫ Identifiers for referencing these representations
3.4.1.1 Class Construct• Similar to
▫“Class” in OO terminology▫“Table” in relational DB terminology
• Group or set of objects with similar properties or characteristics (explicit or implicit) in common
• General statements can be made that apply to all members of the class
• Examples▫Food▫Menu Item▫Person
3.4.1.2 Property Construct•Similar to
▫“Accessor method” in OO terminology▫“Columns” or “fields” in relational DB
terms•Binary association that relates an object
(instance) to a value•Examples
▫Price▫Size
•Unlike OO accessors, properties can be associated with multiple unrelated classes!
3.4.1.3 Individuals• Similar to
▫ “Objects” in OO terminology▫ “Rows” or “records” in relational DB terminology
• Represent class object instances in the domain▫ Physical things▫ Virtual concepts
• Unlike objects, Individuals have no functionality• Examples
▫ KnightOwlRestaurant▫ Order456
• Difference b/w individuals & classes not always clear
• Literal values (“1”, “A”) are special case of individuals
3.4.2 Relating Constructs• Need to describe
relationships between building blocks
• “is an instance of”▫ Individual to Class
• “has value for”▫ Individual to Property
• Restrictions▫ Between Class and
Property
3.4.2.1 Relate Individuals & Classes• Individuals are members
of classes
• “Membership” or “is an instance of” relationship
• Must be explicitly stated
• Examples▫ “KnightOwlRestaurant” is
an instance of “Restaurant” class
▫ “Mark” is an instance of “Person” class
3.4.2.2 Relate Individuals & Properties• Individuals have attributes
described by properties
• “has value for” relationship
• Example▫ “KeyLimePie” individual
has value “$2” for the property “price”
▫ “Mark” individual has value “34” for the property “age”
3.4.2.3 Relate Classes & Properties•Classes can restrict use of
Properties in individuals▫“IsBrotherOf” property range restricted to
“Male”s•Properties can be used to define Classes
by defining membership in the class▫Individual is member of class “Boy” iff
Individual is in “Male” class and “Age” property value <= 18.
•Restrictions can constrain Property values▫To be of a certain class (range)▫To only describe particular classes
(domain)
3.4.3 Semantic Relationships in Blocks •Must be able to describe semantic
relationships within classes, properties, and individuals
•Synonymy•Antonymy•Hyponymy•Meronymy
3.4.3.1 Synonymy Relation• Connects concepts with similar meaning
▫equals() in Java – same meaning, different instance
• Stricter form is equivalence (identical)▫== in Java – same instance
• Class to Class▫Noodles & Pasta; Soda & Pop
• Instance to Instance▫Knight Owl Restaurant & franchiseProperty123
• Property to Property▫Cost & Price
• Allows merging concepts & linking heterogeneous knowledge bases
=
3.4.3.2 Antonymy Relation
•Opposite meaning•Stricter form is disjointness•Establishes dichotomy of meaning b/w
terms•Class to Class
▫Regular Price Menu Item & Sale Price Menu Item
•Instance to Instance•Property to Property
≠
3.4.3.3 Hyponymy Relation• Specialization & generalization• Creates taxonomic hierarchies• Also called
▫ “is-a”▫ “inheritance”▫ “subsumption”
• Transitive downward• Better for permanent relationships• Class to Class
▫ Spaghetti “is-a” Pasta▫ New York Style Pizzeria “is-a” Italian Restaurant “is-a”
Restaurant• Property to Property
▫ salePrice “is-a” price
Δ
Meronymy/Hyponymy Relation• Aggregation & composition• Also called
▫ “part-of”▫ “component of”
• Mereology (part-whole theory)• Holonymy (whole-part theory)• Closely related to “ownership”• Transitive downward• Class to Class
▫ Meatball “part-of” Spaghetti and Meatballs Dish▫ Fork “part-of” Place Setting
• Individual to individual▫ Drink Order 321 “part-of” Restaurant Bill 789
3.4.4 Semantics Summary• Building Blocks • Relationships
Construct Description
A group or set of individual objects with similar characteristics
Associates attrib/value pairs with individuals, restricts classes
Represents a specific instance object of a class
Functionality Relationship Summary
Relating blocks to each other
Individuals to Classes
Membership
Individuals to Properties
Attribute values
Classes to Properties
Restrictions
Describing relationships
Synonymy Similarities
Antonymy Differences
Hyponymy Specialization
Meronymy Part/whole
Holonymy Whole/Part
3.5 Ontology Languages
•Formal, parseable, & usable by software•Define semantics in context-independent
way•Support some level of logic expression•OWL based on:
▫Frame-based systems▫Description logics
3.5.1 Frame-based Systems
•Modeling primitives called “frames” (classes)
•Properties (attributes) are called “slots”•Property values are called “fillers”•Same slot name usable with different
classes▫Can specify different range & value
restrictions
3.5.2 Description Logics (DLs)•Modeling primitives called “concepts”
(classes)•Properties (attributes) are called “roles”•DLs also called “terminological logics” or
“concept languages”•Balance expressiveness with
“decidability”▫Whether software can reach a conclusion
or not•DL concepts defined by their objects’
membership constraints▫Used to automatically derive classification
taxonomies (hierarchies)
3.5.2 Descriptions Logics cont’d•DLs can specify
▫Class constructors▫Property constructors▫Axioms relating classes & properties
•Allow composite descriptions▫E.g. restrictions on relationships between
objects•Use first-order logic•Still decidable•Support efficient inferencing
3.6 Ontologies Summary
•Various definitions (AI, Gruber, OWL)•Purposes
▫Communicate specification of domain▫Declare explicit semantics▫Support information sharing
•Different types; taxonomies most common•Divided into Tbox & Abox
▫Tbox: schema, definitions of concepts▫Abox: records, defintions of
individuals/objects
3.6 Ontologies Summary cont’d•Building blocks
▫Class, Property, Individual•Relationships between different block
types▫Membership, Attribute Values, Restrictions
•Relationships between same block types▫Synonomy, Antonymy, Hyponymy,
Meronymy, Holonymy•Ontologies described using formal
languages
Chapter 4
4.1 OWL Features
•Primary goals▫Intuitive for humans, minimal investment▫Expressive, with explicit semantics for
software•Can define and/or extend ontologies•Supports scalability (needs some work)•XML-based annotations•Makes statements/assertions about
classes, properties, & individuals•Additional facts derived via inferencing
4.2 Layered Architecture
Applications }Implementation Layer
Ontology Languages (OWL Full,
OWL DL, and OWL Lite)}Logical Layer
RDF Schema Individuals }Ontological Primitive Layer
RDF and RDF/XML }Basic Relational Language Layer
XML and XMLS Datatypes }Transport/Syntax Layer
URIs and Namespaces }Symbol/Reference Layer
4.4 OWL Introduction Summary
•Web Ontology Language (OWL)▫Defined by the W3C▫Used to make statements about
Classes Properties Individuals
▫Designed as a layered architecture built on URIs & Namespaces XML & XMLS RDF & RDFS
Backup – Entire slide set
OWLRepresenting Information Usingthe Web Ontology Language
Section 1
Section 1•Chapter 1: Historical Web
▫Web history, context, features, & shortcomings
•Chapter 2: Semantic Web▫Challenges, requirements, & solutions
•Chapter 3: Ontologies▫Concepts, purposes, relationships, features,
& languages•Chapter 4: OWL Introduction
▫OWL language, layered architecture, & supporting technologies
Chapter 1
1 Current Web
•Publishing medium•Dominated by HTML
▫Hyper Text Markup Language•Pages accessible using URLs
▫Uniform Resource Locators▫http://www.w3.org/
•Supports human readers using browsers
1.1 Current Web History
•Internet infrastructure created by DARPA•Mostly text-based (telnet, ftp, gopher)•1992: Tim Berners-Lee/CERT developed
▫HTML & HTTP (Hyper Text Transfer Protocol)
▫Web browser (Mosaic)•Allows anyone to publish structured
documents connected by hyperlinks•Combined with TCP/IP and XML
(eXtensible Markup Language) to create “killer app”
1.2 Current Web Characteristics
•Features
•Benefits
•Applications
1.2.1 Current Web Features
•Diverse•Document-centric•Virtual repository of information•No controlling authority•Managed by open standards from W3C
▫World Wide Web Consortium•Intended for human access & reading
1.2.2 Current Web Benefits
•Superior to private networks•Transactions are cheaper (self-service)•Cheap to communicate world-wide•Created online communities
▫Open-source movement – free high-quality tools
▫Countless online forums
1.2.3 Current Web Applications
•Most content designed for humans•Variety of purposes
▫E-commerce▫Education▫Financial services▫Auctions▫Music
•Many sites use generated HTML & XML generated from databases
1.3 The Web is Not Enough
•Not enough structure to support computer processing of content
•No way to connect information to enable complex queries
•HTML too focused on format/display•Need to add markup to explain meaning
(semantics) •Semantics will enable automated
interpretation of structured web content
1.3.1 Information Structure
•HTML documents▫Semi-structured formatting▫Unstructured text
•Natural Language Processing (NLP)▫Improving, but impractical on a large scale
•Structured database information must be shared in a computer-parseable maner
•Goal: allow automated software agents to mine the web, creating new functionality
1.3.2 Finding Requires Metadata “Find the cheapest Key lime pie within 5 miles.”
• Keyword-based search engines▫ Find pages that might contain desired content▫ Don’t provide answers to questions…the goal!▫ Have to find local restaurants, then look at their
menus• Query engines aim to answer questions
▫ Should be able to filter restaurants within 5 miles, access menus, compare prices, get answer
▫ Show how answer gotten from reliable sources
1.3.3 Semantics Must Be Explicit
•Providing semantic information explicitly in documents enables software to:▫Manipulate information (filter, summarize)▫Infer new facts (inference)▫Link multiple distributed information
representations (semantic join)
1.4 Current Web Summary
•Current Web▫Document-centric▫Focused on humans using browsers▫Insufficient for automated data processing
•New technologies needed▫Structure information for automated
processing▫Improve searches▫Link disparate data sources with each
other•The Semantic Web!
Chapter 2
2 Semantic Web Introduction
•Web information representation challenges
•Requirements for a solution
•Semantic Web concepts that satisfy those requirements
2.1 Web Information Representation Challenges
•Increased Need for Information Representation
•Ambiguous Human Descriptions
•Software Demands for Specificity
2.1.1 Information Representation
•Volume of information increasing exponentially
•User expectations of the Internet also growing
•To satisfy expectations, we need more than just HTML, XML & databases
2.1.2 Ambiguous Descriptions•Many human information formats
▫Specialized domains with unique terminology
▫Regional language differences▫Many sublanguages within communities▫Difficult to get consensus
•Language agreement impossible•Meta-language agreement possible
▫Language to express language•We need a language that can represent
information from many domains
2.1.3 Demands for Specificity
•Computers need information to be▫Structured▫Consistent▫Well-formed▫Logical
2.2 Requirements for a Solution
•Minimize Human Investment
•Satisfy Computer Requirements
•Compromise between these goals
2.2.1 Minimize Human Investment
•Information Representation Producers
•Information Representation Consumers
•Requirements common to both
2.2.1.1 Representation Producers•Provide content from existing sources•Aim to generate information
representations▫Quickly▫Effectively▫Inexpensively
•Represent data using natural models that are▫Extendable▫Versionable▫Configuration-managed
2.2.1.2 Representation Consumers•Aim to create software to
▫Parse information▫Interpret information▫Manipulate information
•Software should be able to▫Combine information from different
domains▫Use others’ data without needing to
understand the underlying data model▫Reduce human intervention
2.2.1.3 Requirements Common to Both
•Solution must be▫Inexpensive▫Easy to implement▫Intuitive▫Evolutionary, not revolutionary▫Compatible with existing web standards
2.2.2 Satisfy Computer Requirements
•Structured distributed representations to enable applications
•Supporting language
2.2.2.1 Structured Representations
•Computers need▫Consistently structured information
collections▫Inference rules to conduct automated
reasoning▫Representations formal enough to detect
inconsistencies and errors▫Network-distributed information to support
scalability
2.2.2.2 Supporting Language
•Need a tagged markup language to provide▫Syntax
Language format rules; open & vendor-neutral
▫Semantics Meaning of concepts; formal, finite, &
extensible▫Expressiveness
Richness; able to express concepts & relationships
Completeness, correctness, & efficiency (hardest!)
▫Standards Common language for all
2.2.3 Compromise
•Must balance need for structure with need for human-friendly data representations▫True natural language processing not yet
ready▫Humans don’t like to process raw
structured data•Proposed solution
▫Humans must augment content with markup
▫Must show an ROI payoff for extra effort
2.3 Semantic Web to the Rescue•Next evolutionary generation of the web
▫Structured information representations provide explicit meaning
▫Information “marked up” according to language standards
▫Software provides new functionality by interpreting, exchanging, & processing meaning
•Technologies focus on information representations tied to explicit meaning
2.3.1 Semantic Web History•Term coined by Sir Tim Berners-Lee•US Dept of Defense/DARPA created
DAML▫DARPA Agent Markup Language▫Helped define critical concepts
•European Union created OIL▫Ontology Interface Layer▫Combined with DAML to create DAML+OIL
•W3C built on DAML+OIL to create OWL▫Web Ontology Language (yes, it’s out of
order)▫First draft approved February 2004
2.3.2 Semantic Web Vision
•Next generation of the web•Vast object-oriented integrated
knowledge base that can be accessed and inferenced via machine-understandable schemas
•Transparent to the end-user•Link documents and the information in
them•Leverage the current web infrastructure•Reduce the cost of performing tasks
2.3.3 Populating the Semantic Web•Developing representation standards
▫Scope the domain/analyze requirements▫Define terms and relationships▫Encode vocabulary & relationships
(ontology)▫Publish representation on servers
•Requires significant up-front effort, but•Yields greater returns than current
solutions•Cost reduces as reuse grows
2.3.4 Use Cases
•Tactical level functionality▫Lower-level functions & basic operations▫Behind the scenes
•Strategic applications▫Higher-level compositions of tactical
features▫Provide more complex functionality▫Customer-facing
2.3.4.1 Tactical Services
•Describe distributed information▫Harvest content, process, & exchange
results•Support queries
▫Answer questions & explain reasoning•Support searching
▫Find information based on meaning, not keywords
•Support inferring▫Drawing conclusions from explicit facts▫Reduces size & complexity of knowledge
bases
2.3.4.2 Strategic Applications•Vertical applications
▫Provide specialized services to a particular domain
▫E-commerce (B2B, B2C)•Agent software
▫Autonomous; mobile; architecture-independent
▫Find & interpret information, act, report results
•Information management▫Migrate intelligence from the software to
the data▫Provide new functionality without
modifying code▫Integrate repositories
2.3.5 Appropriate Applications
•Semantic web applications appropriate to:▫Publish content for both humans and
computers▫Share information without understanding
model▫Inferring new facts & joining information
sources•Characteristics of good candidate
domains:▫Well-understood but dynamic domain▫Heterogeneous information sources▫Existing information interchange
requirements•Not suited to binary data, e.g. image
processing
2.4 Semantic Web Intro Summary•Existing challenges
▫Humans want information in readable formats
▫Computers need structured formats▫Solution must minimize human investment,
but meet computer needs•Semantic web is the solution
▫Builds on the existing web▫Supplies new information representation
features▫Presents information understandable to
both
Chapter 3
3 Ontologies Enable the Semantic Web•Ontology definitions
•Development issues
•Description methods
•Ontology features
•Language issues
3.1 Ontology Definitions•Historical definition
▫Studies of the science of being, and the nature and organization of reality
▫Definitive classifications of objects & their relationships
•Other definitions▫Computer science definition▫Types of ontologies▫Gruber definition▫OWL-specific ontology definitions
3.1.1 Computer Science Definition• Popularized by AI
community • Tbox
▫ Terminogical components▫ Equivalent to “schema”▫ Define concepts▫ Semantic Web equivalent
Ontology• Abox
▫ Assertional components▫ Equivalent to “records”▫ Assert facts▫ Semantic Web equivalent
Individuals
3.1.2 Types of Ontologies• Many types
▫Domain ontologies▫Metadata ontologies (Dublin Core)▫Method/task ontologies
• Many ways to classify ontologies▫Formality▫Regularity▫Expressiveness
• Simplest ontology: Taxonomy▫Hierarchy of concepts related with IS-A
relationship▫Can’t express complex relationships
3.1.3 Gruber Definition
• “Formal specification of a conceptualization” – T. Gruber
• An ontology is a▫ Formally-described▫ Machine-readable▫ Collection of terms &
their relationships▫ Expressed in a language▫ Stored in a file
3.1.4 OWL-Specific Ontology Def’n•Web Ontology Language (OWL) ontology
▫“An OWL-encoded, web-distributed vocabulary of declarative formalisms describing a model of a domain”
•Domain▫A specific subject area or area of
knowledge▫Typically the focus of a particular
community of interest•Encode a model of the domain, not all of it
3.2 Ontology Features
•Communicate a common understanding of a domain
•Declare explicit semantics
•Make expressive statements
•Support sharing of information
3.2.1 Domain Understanding
•Provided by communities of interest▫Example: restaurant association describes
relationships between food items•Ontology formally documents one
common understanding of a domain▫Reduces misunderstanding
•Shared and common understanding communicated between humans and software systems
3.2.2 Explicit Semantics
•Semantics▫Formal descriptions of terms and
relationships▫Traditionally coded into the software or
schema▫Document concepts using modeling
primitives and semantic relationships▫Make assumptions explicit▫Reduce ambiguity▫Enable interoperability
•Must be described formally to be processed
3.2.3 Expressiveness
•“Extensiveness” of the ontology•Must be expressive enough to
▫Represent formal semantics▫Have reasoning properties to support
inferencing•Support canonical granular
representations•Limited to keep reasoning
▫Decidable▫Scaleable
3.2.4 Sharing Information
•OWL-compliant software can▫Manipulate information internally▫Interoperate with other software▫Do semantic mapping between information
sources
•Need to have a shared language and access to information
3.3 Ontology Development Issues• Authoring ontologies
▫Can be developed by anyone, but▫Better if developed by consensus-based
standards development groups▫Vertical ontologies describe a domain▫Horizontal ontologies span domains and
describe basic concepts• Separating ontologies from individuals
▫Usually a good idea▫Sometimes not possible
• Committing to an ontology▫Makes applications easier to understand,
modify, reuse
3.4 Describing Semantics
•Defining information representation building blocks
•Describing relationships between building blocks
•Describing relationships within building blocks
3.4.1 Building Blocks• Three basic blocks
▫ Class constructs▫ Property constructs▫ Individual constructs
• Together, they describe a model of a domain
• Each type requires▫ A computer-
understandable representation
▫ Identifiers for referencing these representations
3.4.1.1 Class Construct• Similar to
▫“Class” in OO terminology▫“Table” in relational DB terminology
• Group or set of objects with similar properties or characteristics (explicit or implicit) in common
• General statements can be made that apply to all members of the class
• Examples▫Food▫Menu Item▫Person
3.4.1.2 Property Construct•Similar to
▫“Accessor method” in OO terminology▫“Columns” or “fields” in relational DB
terms•Binary association that relates an object
(instance) to a value•Examples
▫Price▫Size
•Unlike OO accessors, properties can be associated with multiple unrelated classes!
3.4.1.3 Individuals• Similar to
▫ “Objects” in OO terminology▫ “Rows” or “records” in relational DB terminology
• Represent class object instances in the domain▫ Physical things▫ Virtual concepts
• Unlike objects, Individuals have no functionality• Examples
▫ KnightOwlRestaurant▫ Order456
• Difference b/w individuals & classes not always clear
• Literal values (“1”, “A”) are special case of individuals
3.4.2 Relating Constructs• Need to describe
relationships between building blocks
• “is an instance of”▫ Individual to Class
• “has value for”▫ Individual to Property
• Restrictions▫ Between Class and
Property
3.4.2.1 Relate Individuals & Classes• Individuals are members
of classes
• “Membership” or “is an instance of” relationship
• Must be explicitly stated
• Examples▫ “KnightOwlRestaurant” is
an instance of “Restaurant” class
▫ “Mark” is an instance of “Person” class
3.4.2.2 Relate Individuals & Properties• Individuals have attributes
described by properties
• “has value for” relationship
• Example▫ “KeyLimePie” individual
has value “$2” for the property “price”
▫ “Mark” individual has value “34” for the property “age”
3.4.2.3 Relate Classes & Properties•Classes can restrict use of
Properties in individuals▫“IsBrotherOf” property range restricted to
“Male”s•Properties can be used to define Classes
by defining membership in the class▫Individual is member of class “Boy” iff
Individual is in “Male” class and “Age” property value <= 18.
•Restrictions can constrain Property values▫To be of a certain class (range)▫To only describe particular classes
(domain)
3.4.3 Semantic Relationships in Blocks •Must be able to describe semantic
relationships within classes, properties, and individuals
•Synonymy•Antonymy•Hyponymy•Meronymy
3.4.3.1 Synonymy Relation• Connects concepts with similar meaning
▫equals() in Java – same meaning, different instance
• Stricter form is equivalence (identical)▫== in Java – same instance
• Class to Class▫Noodles & Pasta; Soda & Pop
• Instance to Instance▫Knight Owl Restaurant & franchiseProperty123
• Property to Property▫Cost & Price
• Allows merging concepts & linking heterogeneous knowledge bases
=
3.4.3.2 Antonymy Relation
•Opposite meaning•Stricter form is disjointness•Establishes dichotomy of meaning b/w
terms•Class to Class
▫Regular Price Menu Item & Sale Price Menu Item
•Instance to Instance•Property to Property
≠
3.4.3.3 Hyponymy Relation• Specialization & generalization• Creates taxonomic hierarchies• Also called
▫ “is-a”▫ “inheritance”▫ “subsumption”
• Transitive downward• Better for permanent relationships• Class to Class
▫ Spaghetti “is-a” Pasta▫ New York Style Pizzeria “is-a” Italian Restaurant “is-a”
Restaurant• Property to Property
▫ salePrice “is-a” price
Δ
Meronymy/Hyponymy Relation• Aggregation & composition• Also called
▫ “part-of”▫ “component of”
• Mereology (part-whole theory)• Holonymy (whole-part theory)• Closely related to “ownership”• Transitive downward• Class to Class
▫ Meatball “part-of” Spaghetti and Meatballs Dish▫ Fork “part-of” Place Setting
• Individual to individual▫ Drink Order 321 “part-of” Restaurant Bill 789
3.4.4 Semantics Summary• Building Blocks • Relationships
Construct Description
A group or set of individual objects with similar characteristics
Associates attrib/value pairs with individuals, restricts classes
Represents a specific instance object of a class
Functionality Relationship Summary
Relating blocks to each other
Individuals to Classes
Membership
Individuals to Properties
Attribute values
Classes to Properties
Restrictions
Describing relationships
Synonymy Similarities
Antonymy Differences
Hyponymy Specialization
Meronymy Part/whole
Holonymy Whole/Part
3.5 Ontology Languages
•Formal, parseable, & usable by software•Define semantics in context-independent
way•Support some level of logic expression•OWL based on:
▫Frame-based systems▫Description logics
3.5.1 Frame-based Systems
•Modeling primitives called “frames” (classes)
•Properties (attributes) are called “slots”•Property values are called “fillers”•Same slot name usable with different
classes▫Can specify different range & value
restrictions
3.5.2 Description Logics (DLs)•Modeling primitives called “concepts”
(classes)•Properties (attributes) are called “roles”•DLs also called “terminological logics” or
“concept languages”•Balance expressiveness with
“decidability”▫Whether software can reach a conclusion
or not•DL concepts defined by their objects’
membership constraints▫Used to automatically derive classification
taxonomies (hierarchies)
3.5.2 Descriptions Logics cont’d•DLs can specify
▫Class constructors▫Property constructors▫Axioms relating classes & properties
•Allow composite descriptions▫E.g. restrictions on relationships between
objects•Use first-order logic•Still decidable•Support efficient inferencing
3.6 Ontologies Summary
•Various definitions (AI, Gruber, OWL)•Purposes
▫Communicate specification of domain▫Declare explicit semantics▫Support information sharing
•Different types; taxonomies most common•Divided into Tbox & Abox
▫Tbox: schema, definitions of concepts▫Abox: records, defintions of
individuals/objects
3.6 Ontologies Summary cont’d•Building blocks
▫Class, Property, Individual•Relationships between different block
types▫Membership, Attribute Values, Restrictions
•Relationships between same block types▫Synonomy, Antonymy, Hyponymy,
Meronymy, Holonymy•Ontologies described using formal
languages
Chapter 4
4 OWL Introduction
•OWL Features
•Semantic Web’s Layered Architecture
4.1 OWL Features
•Primary goals▫Intuitive for humans, minimal investment▫Expressive, with explicit semantics for
software•Can define and/or extend ontologies•Supports scalability (needs some work)•XML-based annotations•Makes statements/assertions about
classes, properties, & individuals•Additional facts derived via inferencing
4.2 Layered Architecture
Applications }Implementation Layer
Ontology Languages (OWL Full,
OWL DL, and OWL Lite)}Logical Layer
RDF Schema Individuals }Ontological Primitive Layer
RDF and RDF/XML }Basic Relational Language Layer
XML and XMLS Datatypes }Transport/Syntax Layer
URIs and Namespaces }Symbol/Reference Layer
4.2 Layered Architecture cont’d• Layers illustrate rough
dependencies▫ Each layer uses features
of lower layers• Implementation Layer
▫ Provides specific applications
• Logical Layer▫ OWL supports formal
semantics and reasoning• Ontological Primitive
Layer▫ RDFS defines vocabulary▫ Individuals defined in
RDF
RDF Schema Individuals
XML and XMLS Datatypes
URIs and Namespaces
Applications
Ontology Languages (OWL Full, OWL DL, and OWL Lite)
RDF and RDF/ XML
4.2 Layered Architecture cont’d• Relational Language Layer
▫ RDF’s simple data model & syntax for making statements
▫ Serialized as RDF/XML or N-triples
• Transport/Syntax Layer▫ Define primitive
datatypes▫ Provide encoding format
• Symbolic/Reference Layer▫ Identify and reference
classes, properties, and individuals
RDF Schema Individuals
XML and XMLS Datatypes
URIs and Namespaces
Applications
Ontology Languages (OWL Full, OWL DL, and OWL Lite)
RDF and RDF/ XML
4.3 Technology Support for Layers•Symbol/Reference Layer
▫Provides identifiers & references to objects described in ontologies and instance files
•Transport/Syntax Layer▫XML used to serialize OWL syntax▫XMLS defines standard datatypes
•Basic Relational Layer▫RDF makes statements using
Attribute/Value pairs to describe objects
4.3 Tech Support for Layers, cont’d• Ontological Primitive Layer
▫RDFS provides basic vocabulary describing Classes and subclasses Properties and subproperties
▫Instances & property values specified by RDF & XMLS
• Logical Layer▫OWL dialects (Full, DL, Lite) enhance RDFS
• Implementation Layer▫Applications built using OWL
• Additional layers being considered for rules & trust
4.4 OWL Introduction Summary
•Web Ontology Language (OWL)▫Defined by the W3C▫Used to make statements about
Classes Properties Individuals
▫Designed as a layered architecture built on URIs & Namespaces XML & XMLS RDF & RDFS
top related