The Knowledge Cartography – A New Approach to Reasoning over =: S

Download Report

Transcript The Knowledge Cartography – A New Approach to Reasoning over =: S

=: SOFSEM 06 :=
The Knowledge Cartography –
A New Approach to Reasoning over
Description Logics Ontologies
Krzysztof Goczyła, Teresa Grabowska,
Wojciech Waloszek, Michał Zawadzki
January 21-27, 2006, Měřín, Czech Republic
PIPS
Upper layers
Knowledge
KMS
KaSeA RDB
KaSeA
RDB
Data
Internet
© 2006 KMG @ GUT
#2
Motivations
• PIPS manages knowledge about:
– ilnesses,
– drugs,
– alergies,
– diets,
–…
• Required data are gathered from many
internet sources; the number of
information is very large.
© 2006 KMG @ GUT
#3
Knowledge Cartography
• Main assumptions:
– Terminology is updated so rarely that it might be
considered constant in time.
– A knowledge base is queried much more often
than updated (new individual assertions).
– A knowledge base should be able to hold and
efficiently process information about large
numbers of individuals.
© 2006 KMG @ GUT
#4
Knowledge Cartography
• Developed to:
– simplify reasoning on numerous objects
(individuals),
– simplify storage of inferred results concerning
numerous individuals.
• Processes ontological information in
accordance to Description Logics and
Semantic Web standard.
© 2006 KMG @ GUT
#5
Terminology – an example
• Terminology is described by a set of
statements - axioms.
• Suppose, we want to describe the world of
people and their pets:
– There are no other creatures except People,
Dogs and Cats,
– People, Dogs and Cats are disjoint,
– Dogs and Cats are CreaturesHavingTails,
– People and Cats can be
CreaturesTakingCareOfHygiene.
© 2006 KMG @ GUT
#6
Map of concepts
There are no other creatures except People, Dogs and Cats
People  Dogs  Cats  
© 2006 KMG @ GUT
#7
Map of concepts
People, Dogs and Cats are disjoint
People  Dogs  
People  Cats  
Cats  Dogs  
© 2006 KMG @ GUT
#8
Map of concepts
Dogs and Cats are CreaturesHavingTails
CreaturesHavingTails  Dogs  Cats
© 2006 KMG @ GUT
#9
Map of concepts
People and Cats can be CreaturesTakingCareOfHygiene
CreaturesTakingCareOfHygiene  People
CreaturesTakingCareOfHygiene  Cats
© 2006 KMG @ GUT #10
Map of concepts
© 2006 KMG @ GUT #11
Map of concepts – signatures
1
2
3
4
5
11000

01000
00001

00010
00111

11111
© 2006 KMG @ GUT #12
Terminological queries
Must every person take care of hygenie?
11000
NO
10011
Is a creature without a tail a person?
11000
YES
11000
Is a creature which have a tail and takes care of hygenie a cat?

00100
YES
00110
© 2006 KMG @ GUT #13
Individuals’ signatures
• Individuals’ signatures describe what we know
and what we do not know about individuals
Source 1: Fred has a tail
00111
Is Fred a person?
00111
11000
NO
Is Fred a dog?
00111
00001
MAYBE
© 2006 KMG @ GUT #14
We know something more…
Source 2: Fred takes care of hygenie
00111
01100
,
00100
Is Fred a dog?
00100
00001
NO
Is Fred a cat?
00100
YES
00110
© 2006 KMG @ GUT #15
Restrictions
• Role constructors in the form of R.C and
R.C are treated as concepts
• Such constructors must be explicitly defined
in ontology.
© 2006 KMG @ GUT #16
Deployment in PIPS
• Cartographic approach has been succesfully
deployed in the first release of PIPS system,
• Allowed for achieving short (<1 s) response
times for queries in knowledge bases with
very large (>10000) number of individuals
(sublinear scalability).
© 2006 KMG @ GUT #17
Future work
• Compression of signatures and hierarchical
signatures,
• Introducing signatures not only for concepts
but also for roles (relations) between
individuals.
• Extending expresiveness of KaSeA in order
to support complex queries (cardinality
constraints, functional, symetric and
transitive roles),
© 2006 KMG @ GUT #18
Thank you!
=: SOFSEM 06 :=
© 2006 KMG @ GUT #19