Upload
eswcsummerschool
View
88
Download
1
Embed Size (px)
DESCRIPTION
Citation preview
© Copyright 2007 STI - INTERNATIONAL www.sti2.org
An Introduction to The Semantic Web: a knowledge
modelling perspective John Domingue
President STI International
www.sti2.org
Caveat
• Not an historical overview • Singular perspective of some underlying
motivations and issues related to the Semantic Web
www.sti2.org
MOTIVATION & APPLICATION SCENARIO
www.sti2.org
Agenda
• Building K-HAL v1.0 – Influences – Ontology – Conceptualisation – Knowledgebase – Critique
• Building K-HAL v2.0 – Ontologies – Data – Getting help
• Building K-HAL v3.0 • Conclusions
www.sti2.org
BUILDING K-HAL V1.0
Spacecraft piloting and navigation only – the reasoning and knowledge parts
5/12/2007 - Vienna
www.sti2.org
INFLUENCES
www.sti2.org
8/59
Cognitive Psychology
www.sti2.org
9/59
Semantic Network Collins and Quillian 1967
www.sti2.org
10/59
Knowledge Level (Allen Newell, 1982)
Knowledge Level
Symbol Level
Physical Level
.
.
.
.
.
www.sti2.org
11/59
Knowledge Level
Symbol Level
Physical Level
.
.
.
.
.
Knowledge
Goals
Principle of Rationality
Agent
Observer
Knowledge Level (Allen Newell, 1982)
www.sti2.org
ONTOLOGY
www.sti2.org
Informal Ontology Explanation
• Used to structure knowledge • Facilitates interoperability • Formal explicit shared
conceptualisation of a domain • A set of concepts, relationships
and individuals over which there is an agreed consensus
www.sti2.org
14/59
Ontology Construction
www.sti2.org
CONCEPTUALISATION
www.sti2.org
16/59
Let’s talk to some smart people…
www.sti2.org
17/59
Conceptualisation
www.sti2.org
18/59
Conceptualisation
1 3
5
7 9
2
4 6
8
www.sti2.org
19/59
Re-conceptualisaiton
5
2
4 6
8 1
7
3
9
www.sti2.org
20/59
Re-conceptualisaiton
5
2
4 6
8 1
7
3
9
www.sti2.org
21/59
Re-conceptualisaiton
5
2
4 6
8 1
7
3
9
www.sti2.org
22/59
Re-conceptualisaiton
5
2
4 6
8 1
7
3
9
www.sti2.org
23/59
Re-conceptualisaiton
5
2
4 6
8 1
7
3
9
www.sti2.org
24/59
Re-conceptualisaiton
5
2
4 6
8 1
7
3
9
www.sti2.org
25/59
Re-conceptualisaiton
5
2
4 6
8 1
7
3
9
www.sti2.org
26/59
Re-conceptualisaiton
5
2
4 6
8 1
7
3
9
www.sti2.org
27/59
K-HAL v 1.0 Ontology (small portion)
Space Ship
Has Component
Rocket Generates Thrust
Concepts Relations
Star
Engineered Artifact
Celestial Body
Planet Asteroid
Crew
Human Crew
Onboard AI
Agent
Has Name
Has Mass
Has Volume
www.sti2.org
K-HAL V1.0 KNOWLEDGE BASE
www.sti2.org
29/59
K-HAL v 1.0 Ontology (small portion)
Space Ship
Has Component
Rocket Generates Thrust
Concepts Relations
Star
Engineered Artifact
Celestial Body
Planet Asteroid
Crew
Human Crew
Onboard AI
Agent
Has Name
Has Mass
Has Volume
www.sti2.org
30/59
K-HAL v 1.0 Ontology/KB (small portion)
Space Ship
Has Component
Rocket Generates Thrust
Concepts Relations
Star
Engineered Artifact
Celestial Body
Planet Asteroid
Crew
Human Crew
Onboard AI
Agent
Has Name
Has Mass
Has Volume
HAL
Dave Boorman
The Sun Jupiter
www.sti2.org
31/59
Knowledge Base
www.sti2.org
32/59
CYC Open CYC v 4.0 239,000 concepts and 2,093,000 facts
www.sti2.org
33/59
K-HAL v 1.0
Ontology
Knowledge Base
Input/Output Vision system
Speech Generation Speech Understanding
User Reasoner
www.sti2.org
What’s wrong with K-HAL v 1.0?
• ????
www.sti2.org
BUILDING K-HAL V2.0
5/12/2007 - Vienna
www.sti2.org
36/59
Underlying Principles
Be Lazy
Be kind and share
www.sti2.org
K-HAL V2.0 ONTOLOGIES
www.sti2.org
38/59
Why work when you can Google?
www.sti2.org
39/59
DBpedia
www.sti2.org
41/59
Astronomical Objects
www.sti2.org
42/59
Astronomical Ontology (portion)
www.sti2.org
43/59
Space Shuttle Ontology
Aft-Fusulage Air-Sampling-System Airlock Auxilliar-Power-Unit Avionics-System Body-Flap Caution-And-Warning-System Communication Crew-Apparel Crew-Compartment Crew-Equipment Crew-Equipment-Stowage Eletrical-Power-Distribution Eletrical-Power-System Emergency-Egress-Slide Environmental-Control-and-Life-Support-System Exercise-Equipment External-Tank External-Tank-Hardware Food-System-and-Dining Forward-Fuselage Hold-Down-Post Housekeeping Hydaraulic-System Hydarulic-Power-Units
Inflyght-Crew-Escape-System Intertank Landing-Gear-System Liquid-Hydrogen-Tank Liquid-Oxygen-Tank Main-Propulsion-System Microcassete-Recorder Midfuselage Operational-Bioinstrumentation-System Orbial-Manuevering-System Orbital-Maneuvering-System Orbiter-External-Tank-Seperation-System Orbiter-Flight-Crew-Escape-System Orbiter-Passive-Contro-Systeml Orbiter-Purge-Vent-Drain-System Orbiter-Structure OrbiterManufacturingandAssembly Payload-Bay-Doors Personal-Hygiene-Provision Photographic-Equipment Radiation-Equipement Range-Safety-Siystem Range-Safety-Siystem-Tank Reach-Aid Reaction-Control-System Restraints-And-Mobility-Aids
RSB-Descent-and-Recovery Secondary-Emergency-Egress Shuttle Shuttle-Orbiter-Medical-System ShuttleName Sighiting-Aids Sleeping-Provision Solid-Roket-Boosters Sound-Level-Meter Space-Shuttle-Coordinate-System Space-Shuttle-Orbiter-System Space-Transportation-System SRB-Ignition SRB-Rate-Gyro-Assemblies SRB-Separation Thermal-Protection-System Thermal-Protection-System-Orbiter Thrust-Vector-Control Vertical-Tail Water-Spray-Boilers Wicket-Tabs Windows Wing
www.sti2.org
Problems to be resolved (ontology)
• Finding ontologies • Understanding ontologies • Connecting ontologies • Adapting ontologies • Version control • Agility
– New ontologies, changes in used ontologies …
• …….
www.sti2.org
45/59
Building and Using Ontologies
Elena Simperl Tutorial 10:45am Tuesday
www.sti2.org
K-HAL V2.0 DATA
Be Lazy
www.sti2.org
47/59
Solar Objects > 400km Radius
www.sti2.org
48/59
www.sti2.org
49/59
Spacecraft component data
www.sti2.org
Problems to be resolved
• Finding semantic data • Transforming unstructured data to a semantic
format • Transforming structured data to a semantic format • Connecting semantic datasets • Querying/reasoning over connected semantic data • Sharing new data • Agility
– New datasets, changes in used datasets…
www.sti2.org
Linked Data Basics
• Fundamentals of Linked Data: main standards & technology components, motivating application scenario – Barry Norton Tutorial 10:45am today
• Querying Linked Data: SPARQL 101 – Irini Fundulaki Tutorial 2pm today
• Semantic Web languages and standards: RDF, RDFS, SPARQL – Barry Norton & Irini Fundulaki Hands-on: 3:30pm today
www.sti2.org
Publishing and Using Linked Data
• Providing and consuming Linked Data – Maribel Acosta Tutorial 2:30pm Tuesday
• Publishing and consuming Linked Open Data – Maribel Acosta Hands-on 4pm Tuesday
www.sti2.org
Linked Data and the Unstructured World
• Linked Data for NLP – Barry Norton Tutorial Wednesday 10:45am
• Using Linked Data and GATE – Barry Norton & Isabelle Augenstein Wednesday 11:30am
www.sti2.org
54/59
Getting everybody to help…
*picture suggested by Dan Brickley
www.sti2.org
55/59
GalaxyZoo
www.sti2.org
56/59
More Zoos
www.sti2.org
57/59
Crowdsourcing AI
www.sti2.org
Getting help tutorials and hands-on
• Social Semantic Web and crowdsourcing – Elena Simperl Tutorial Wednesday 2pm
• Using Mechanical Turk to solve Linked Data problems – Maribel Acosta Hands-on Wednesday 3pm
www.sti2.org
59/59
K-HAL v 1.0
Ontology
Knowledge Base
Input/Output Vision system
Speech Generation Speech Understanding
User Reasoner
www.sti2.org
60/59
K-HAL v 2.0 Architecture
Linked Open Vocabularies
Linked Open Data
HAL Ontology
HAL Facts in RDF Store
Reasoner
Input/Output Vision system
Speech Generation Speech Understanding
Crowdsourced facts
Corporate data
www.sti2.org
BUILDING K-HAL V3.0 What about processes?
5/12/2007 - Vienna
www.sti2.org
www.sti2.org
63/59
Virtual Choir
www.sti2.org
64/59
Choir Autonomous singers Available online
www.sti2.org
65/59
Conductor Dictates song Common notation Selects performances Edits and mixes
Choir Autonomous singers Available online
www.sti2.org
66/59
Listener Has a desire Has preferences
Conductor Dictates song Common notation Selects performances Edits and mixes
Choir Autonomous singers Available online
www.sti2.org
67/59
www.sti2.org
Conclusions (1/2)
In its current state the Semantic Web/Web of Data facilitates the re-use of ontologies and data • Other problems arise associated with
ontology and data quality, adapting/aligning ontologies and data …
• Good SW/LD practitioners know online ontologies and datasets as a good researcher knows the related literature
Be Lazy
www.sti2.org
Conclusions (2/2)
Releasing ontologies and data • Provides a community benefit for
expected and unexpected uses • Can increase the value of the released
artifacts • May be obligated depending on context
(e.g. if paid for by public funding) • Has associated issues related to
training, quality, privacy, maintenance….
Be kind and share
www.sti2.org
THANKS