Distributed Instance Retrieval over Heterogeneous Ontologies

1
Distributed Instance Retrieval over Heterogeneous
Ontologies
2nd Italian Workshop on Semantic Web Applications
and Perspectives

• Andrei Tamilin(1,2) Luciano Serafini(1)
• (1) ITC-IRST
• (2) DIT - University of Trento

Trento, 16 December, 2005
2
Outline

• Motivation and problem
• Distributed Description Logics (DDLs) with
  individuals
• Reasoning/Querying patterns in DDLs
• Instance retrieval in DDLs
• Implementing

3
Motivation a step back

• Where we were
• Steady ontology proliferation
• Heterogeneity is inevitable
• Problem
• How to interoperate?
• Requirements
• Semantic mappings
• Reasoning support
• Solution
• Distributed Description Logics
• Distributed tableaux
• DRAGO reasoning system

4
Motivation a step forth

• Where we are
• Ontologies are populated
• Populations can be done in heterogeneous domains
• Problem
• How to query such a system?

5
A toy example
ODIT
OUniTn
? Retrieve all personal contacts
andrei.tamilin_at_unitn.it
http//dit.unitn.it/tamilin

• In case of which semantic correspondences
• the query propagation should occur?
• How the individuals should be transformed?

andrei.tamilin_at_dit.unitn.it tamilin_at_itc.it
6
Requirements / Our proposals
Requirements / Our proposals

• Formal framework reflecting conceptual and
  individual heterogeneity
• Extend Distributed Description Logics with
  individuals
• Define a suitable query answering procedure
• Extend Distributed tableau algorithm
• Implement the querying procedure
• Extend DRAGO reasoning system

7
DDLs in a nutshell

• Captures the case of multiple ontologies pairwise
  linked by semantic mappings
• Ontologies correspond to DL knowledge bases
• Mappings correspond to bridge rules

8
DDLs syntax

• DDL is a family of description logics DLii?I
• A bridge rule from i to j is an expression of the
  form
• where X and Y are concepts of DLi and DLj.
• A distributed T-box (DTB) is a pair ?Tii?I,
  Biji?j?I?
• where Bij is a collection of bridge rules from i
  to j

9
DDLs syntax with individuals

• An individual correspondence from i to j is an
  expression of the form
• where x and y are individuals of Ai and Aj.
• A distributed A-box (DAB) is a pair ?Aii?I,
  Ciji?j?I? where Cij is a collection of
  individual correspondences from i to j
• A distributed knowledge base (DKB) is a pair
  ?DTB, DAB?

10
DDLs semantics
DLii?I
Biji?j?I
Tii?I
DTB ?Tii?I, Biji?j?I ?
Aii?I
Ciji?j?I
DTA ?Aii?I, Ciji?j?I ?
Jii?I
riji?j?I
DI ?Jii?I, riji?j?I ?
11
Into-bridge rule
DI
rij(xIi) ? YIj
?Ii
?Ij
Y
X
rij(X)
rij
12
Onto-bridge rule
DI
rij(XIi) ? YIj
rij(X)
?Ii
?Ij
X
rij
Y
13
Individual correspondence
DI
? xIi,yIj ??rij
?Ii
?Ij
rij
y
x
14
Terminological propagation
DTB ?T1, T2, B12?
T1
T2
A
G
isA
H
B
GI2 ? r12(AI1)
r12(BI1) ? HI2
?
15
Assertion propagation
DKB ??T1,A1?, ?T2,A2?, B12,C12?
T1
T2
B
H
isInstanceOf
isInstanceOf
A1
A2
b
h
hI2 r12(bI1)
r12(BI1) ? HI2
?
16
Assertion propagation - II
DKB ??T1,A1?, ?T2,A2?, B12,C12?
T1
T2
B
H
isInstanceOf
isInstanceOf
fij
A1
A2
fij(b)
b
17
Distributed instance retrieval

• Instance retrieval finding all individuals that
  instantiate a given concept
• Both propagation aspects should be taken into
  account

18
D1
Local taxonomy
D3
D2
Local individuals
i2
i1
i3
Retrieve instances of D1 ? i1, i2
19
Bridge rules
Distributed taxonomy (terminological propagation
matters)
D1
D3
D2
Local individuals
i2
i1
i3
Retrieve instances of D1 ? i1, i2, i3
20
Bridge rules
Distributed taxonomy (terminological propagation
matters)
D1
D3
D2
Individual correspondences
Local individuals
i2
i1
i3
Distributed individuals (transformed via
individual correspondences)
i2
Retrieve instances of D1 ? i1, i2, i3, i2
21
Implementation

• On top of DRAGO distributed terminological
  reasoner
• DRAGO is a peer-to-peer network of communicating
  reasoners that handle OWL ontologies coupled with
  C-OWL mappings
• For the instance retrieval the possibility to
  specify instance transformations has been added

22
Conclusions and Outlook

• We addressed the problem of retrieving
  individuals over heterogeneous ontologies which
  are instantiated in semantically related domains
• We extended DDLs framework with individual
  correspondences and discussed how this enables
  the propagation of assertions over ontologies
• Implemented a simple querying prototype on top of
  DRAGO reasoner

23
Thank you