Διαφάνεια 1

Download Report

Transcript Διαφάνεια 1

CHOROS: A Reasoning and Query Engine
for Qualitative Spatial Information
Georgios Christodoulou, Euripides G.M. Petrakis, and Sotirios
Batsakis
Department of Electronic and Computer Engineering,
Technical University of Crete (TUC) Chania, Crete,
1
Motivation
 Qualitative information is expressed without
numerical values using a vocabulary of relationships
 closer to how humans represent and reason about
commonsense knowledge
 It it is possible to deal with incomplete knowledge
 Reasoning over qualitative spatial information is the
problem this work is dealing with.
 Two of the most important aspects of space are
topology and orientation.
2
Topological Relations
 Region Connection Calculus (RCC) abstractly describes regions in a topological space
by means of 8 basic relations:








3
disconnected (DC)
externally connected (EC)
equal (EQ)
partially overlapping (PO)
tangential proper part (TPP)
tangential proper part inverse (TPPi)
non-tangential proper part (NTPP)
non-tangential proper part inverse (NTPPi)
Directional Relations
 Cone-shaped Directional (CSD): relative directional position
between two points in space by means of 9 basic relations:
 north (N)
 north-east (NE)
 east (E)
 south-east (SE)
 south (S)
 south-west (SW)
 west (W)
 north-west (NW)
 identical (O)
4
Qualitative Spatial Reasoning
 Refers to the process of computing new relations
from a set of existing ones and detecting
inconsistencies
 Using some spatial algebra like CSD-9 and
RCC-8
 Relies on a Composition table for each calculus
 Path Consistency
5
RCC-8 Composition
6
Directional Composition
7
SOWL [Batsakis 2011]
 SOWL is a framework for handling spatio-temporal
information:
 An ontology for spatial and temporal concepts.
 A reasoner implemented using SWRL rules and OWL
2.0 constructs (e.g., disjoint properties) ensuring path
consistency.
 A spatio-temporal query language
 The SOWL spatial representation supports both RCC
and CSD calculi.
8
PelletSpatial [Stocker 2009]
 PelletSpatial extends Pellet with qualitative spatial
reasoning over RCC relations.
 Implements two RCC reasoners:
 One implementing translation of RCC relations to
OWL-DL class axioms while preserving their
semantics.
 One operating on the RCC composition table by
implementing a path-consistency algorithm
 Doesn't support directional (CSD) algebra
9
CHOROS Spatial reasoner
 CHOROS extends PelletSpatial to support CSD relations in
addition to RCC relations.
 It implements a path-consistency algorithm based on the
composition tables used in SOWL.
 query answering
 Spatial relations are expressed in RDF/OWL forming an
ontology. A relation is represented as a triple.
 we represent a region as an OWL individual
 Spatial relations are defined as object properties
10
Spatial Representation
 CHOROS provides an RDF/OWL vocabulary for expressing
qualitative spatial relations, with both the CSD and RCC
models. One can use his/her own by defining sub-property
axioms. (e.g., "borders" sub-property of "externally
ConnectedTo")
11
CHOROS Architecture - Components
 Parser: loading ontologies, queries
 Reasoner: consistency checking
 Query Engine: answering queries
 CHOROS separates spatial reasoning from
semantic OWL-DL reasoning.
12
CHOROS Reasoner
 It is realized by means of a path-consistency algorithm ensuring that
computed and existing relations are consistent
 A queue Q keeps track of relations that have to be processed. The
algorithm runs until Q = ∅ or an inconsistency is detected. Q is
initialized with all the defined relations Rij ∊ N
1.
2.
3.
4.
13
We process N to infer all the inverse and equals relations.
We compute the compositional inference Tac ⟵ Rab ⃘ Sbc (a
composition table lookup)
We complete intersections Vac ⟵ Tac ⋂Uac
A relation Rab is path-consistent if the rule Vac ⟵ Uac ⋂ Rab ∘ Sbc
results in V ≠ ∅.
Reasoning Example
 A spatial configuration is formalized in CSD as the following constraint
network:
 house1 N house2
 house2 NW house3
 house1 NE house4
 house4 N house3
 Using the CSD composition table and the path-consistency algorithm,
we can refine the network in the following way:
 house1 N, NW house3
 house1 N, NE house3
 That is, the first house is north of the third which is the intersection of
the above two relations.
14
CHOROS Variations
 CHOROS 0.1 applies over all 9 CSD calculus basic relations.
 CHOROS 0.2 applies to consistency checking over 8 CSD basic
relations ("identical to" is replaced by the owl axiom "sameAs”)
 Multithreading allows two parts of the same program to run
concurrently. We utilize multithreading by launching each
calculi as a separate thread.
 In CHOROS as well as in PelletSpatial, path consistency has O(n3)
worst time complexity (with n being the number of individuals)
15
Experiments
 The "TUC spatial ontology" describes the spatial entities of
the campus of Technical University of Crete
16
TUC Spatial Ontology
17
Reasoning times
18
Conclusions & Future Work
 We presented CHOROS, a qualitative spatial reasoning and query
engine implemented in Java. CHOROS supports both RCC and
CSD models.
 We evaluated possible optimizations of CHOROS (CSD-8,
multithreading) and compare its performance with that of a spatial
reasoner implemented in SWRL.
 Future work includes:
 extending our implementation to support qualitative temporal
reasoning on basic Allen relations
 supporting reasoning beyond the base relations of each calculi (PP as
a disjunction of TPP, NTPP)
19
Thank You
Questions?
20