Upload
percival-hill
View
216
Download
1
Tags:
Embed Size (px)
Citation preview
Clément Troprès - Damien Coppéré 1
Semantic Web
Based on:
-The semantic web
-Ontologies Come of Age
Clément Troprès - Damien Coppéré 2
Plan
Introduction to semantic web
Kwnoledge Representation
Ontologies
Agents
Clément Troprès - Damien Coppéré 3
1. Introduction to semantic web
Today, most of the web contents is designed for human to read
The actual web looks insufficient
The semantic web purpose is to structure the world wide web
Clément Troprès - Damien Coppéré 4
1. Introduction to semantic web
Principles:
1. Each object of the web has a metadata
2. Each metadata is readable by agents and humans
3. Each metadata represents accurately an object
4. Each metadata is available in a common space, readable by agents and humans. The selection of the metadata makes the object avalaible as a resource
Clément Troprès - Damien Coppéré 5
1. Introduction to semantic webThe semantic web architecture
Clément Troprès - Damien Coppéré 6
2. Knowledge representation (1):
Technology which permits computers to access to structured collections of information
System must have sets of inference rules that computers can use to conduct automated reasoning
It has to be linked into a single global system
Clément Troprès - Damien Coppéré 7
2. Knowledge representation (2) :
Traditional systems usually :
- Limit the questions that can be asked
- Become unmanageable
New systems, in contrast, accept paradoxes
- Unanswerable questions are a price that must be paid to achieve versatility.
Clément Troprès - Damien Coppéré 8
2. Knowledge representation (3) :
Two important technologies exist :
- EXtensible Markup Language (XML)
- Resource Description Framework (RDF)
XML :
- Everyone can create their own tags
- It allows to add arbitrary structure to the document
Clément Troprès - Damien Coppéré 9
2. Knowledge representation (4) : RDF :
- Encode in sets of triplets - Each triple being rather like the subject, predicate and object of an elementary sentence identified by URIs - Natural way to describe the vast majority of the data processed by machines - Example : New York has a postal abbreviation which is NY
<rdf:Description rdf:about="urn:states:New%20York"> <"http://purl.org/dc/terms/" :alternative>NY</rdf:Description>
Universal Resource Identifier - Ensure that concepts are tied to a unique definition that
everyone can find on the Web
Clément Troprès - Damien Coppéré 10
3. Ontologies - Introduction
Current web :
It has grown and continues to grow very quickly
Problems to find information you are really looking for
Designed for human perception
Semantic web:Make the web understandable by computers agent
Clément Troprès - Damien Coppéré 11
3. Ontologies - Introduction
How make the web semantic?
- Complete HTML tag (with XML)
- Organize the keywords in each document
- Indexing all the resources of the web (RDF)
- Ontologies
Clément Troprès - Damien Coppéré 12
3. Ontologies - Introduction
We arehere
Clément Troprès - Damien Coppéré 13
3. Ontologies - Introduction
Definition:- In 1993, Gruber propose his definition (which is now the most cited in AI) :
« An ontology is an explicit specification of aconceptualization ». (Gruber T., 1993b)
- In 1997, Borst modified slightly the definition in order to highlight major aspects of this paradigm:
« An ontology is a formal specification of a sharedconceptualization ». (Borst W. N., 1997)
Clément Troprès - Damien Coppéré 14
3. Ontologies - Introduction
Definition:In 1998, these two definitions were only one in the definition of Studer.
« An ontology is a formal, explicit specification of a shared conceptualization ». (Studer R. et al., 1998)
- « Conceptualization » refers to an abstraction of a phenomenon obtained by identifying the concepts appropriate to this phenomenon - « Shared » means that ontology captures consensual knowledge
Clément Troprès - Damien Coppéré 15
3. Ontologies - Introduction « Formal » means that ontology is interpretable by a
machine (machinereadable)
« explicit specification » means that the concepts of ontology and the constraints related to their use are defined in a declaratory way
Ontology has the following characteristics :
1) shared, 2) explicit, 3) formal
Clément Troprès - Damien Coppéré 16
3.Ontologies – Possible representation?
A controlled vocabulary (eg: Catalogs) A glossary (list of terms) Thesauri (synonym relationship…, but not an explicit
hierarchy) Term hierarchies (without true subclass) Strict subclass hierarchies Frames (classes include property information) Value restriction (eg: a price is a number) Logical deduction
A
B
A is a superclass of B
Clément Troprès - Damien Coppéré 17
3. Ontologies – Simple Ontologies
Some of the ways that simple ontologies may be used in practice:
- A controlled vocabulary (beginning of interoperability)- Site organization and navigation support- Expectation setting- Umbrella structures from which to extend content- Browsing support- Search support- Sense disambiguation support
Clément Troprès - Damien Coppéré 18
3. Ontologies – Structural Ontologies
- Consistency checking- Completion- Interoperability support- Support validation and verification testing- Encode entire test suites- Configuration support- Support structured, comparative and customized search- Exploit generalization/specialization information
Clément Troprès - Damien Coppéré 19
3. Ontologies – Implications and Needs
An ontology-based application has two major concerns:
The language
The environment
Clément Troprès - Damien Coppéré 20
3. Ontologies – Implications and Needs (1)
The language:
Simple ontologie: It’s not a real problem (language with subclass and instance relationships)
Structural ontologie: the language must be able to express the entire domain unambiguously (KRSS, KIF, OKBC)
Clément Troprès - Damien Coppéré 21
3. Ontologies – Implications and Needs (2)
Environment:
Ontology tools are needed to analyze, modify and maintain an ontology over time
Many are avalaible commercially
Clément Troprès - Damien Coppéré 22
3. Ontologies – Implications and Needs (3)
Environment – Criterias needed :
- Collaboration and distributed workforce support (share session)
- Platform interconnectivity (able to read and write compatible formats)
- Scale (In terms of size of ontologies, number of simultaneous users)
- Versioning (Able to support many versions of ontology)
Clément Troprès - Damien Coppéré 23
3. Ontologies – Implications and Needs (4)
Environment – Major criteria of performance :
- Security
- Analysis (focus the user’s attention in areas which need modification)
- Lifecyle issues (Support for ontology evolution issues)
- Ease of use (training materials, tutorials…)
- Diverse user support
- Presentation style
- Extensibility (Adapt along with the needs)
Clément Troprès - Damien Coppéré 24
4. Agents
Representing by programs :
- Collect Web content from diverse sources
- Process the information
- Exchange the results with other programs
All agents can work together
Clément Troprès - Damien Coppéré 25
4. Agents (2)
Important facets :
- "Proofs" written in the Semantic Web's unifying language (Proof Markup Language PML)
- Digital signatures used to verify that the attached information has been provided by a specific trusted source
Example of agent : You answer your phone and the
stereo sound which was working is turned down.
Clément Troprès - Damien Coppéré 26
4. Agents (3) You want to buy a car …
An intelligent Agent is going to find your new car- How ? It looks for all cars which corespond to your criterias- Which criteria ?Prices, delivery period, colour… - Where ? On web documents described by semantic standards (proofs, digital signature…)
Travel Agency…
Clément Troprès - Damien Coppéré 27
- Lets anyone express new concepts with minimal effort
- Unifies a logical language
The Semantic Web