An Application of Semantic Web Technologies to Situation Awareness

1
An Application of Semantic Web Technologies to
Situation Awareness
An Application of Semantic Web Technologies to
Situation Awareness

• Christopher J. Matheus
• Versatile Information Systems, Inc.
• Mieczyslaw M. Kokar, Kenneth Baclawski
• Northeastern University
• Jerzy A. Letkowski
• Middle New England College

www.vistology.com
2
Overview

• Situation Awareness
• Ontologies and Rules
• Situation Awareness Assistant SAWA
• A Simple Scenario
• Lessons Learned

3
Our Problem Domain

• Formal yet Practical Applications of Situation
  Awareness
• Situation Awareness (SAW)
• an understanding of whats going on in an
  evolving situatione.g. supply logistics,
  financial markets, battlefields
• involves fusion of object-level data from
  multiple sources into meaningful higher-order
  relations
• highly context dependent and goal directed (i.e.,
  requires domain knowledge)
• Requirements for effective SAW apps
• domain knowledge about relevant objects and their
  properties
• specification of conditions that define
  higher-order relations
• a means for reasoning about time-dependent sensor
  information in the context of the given domain
  knowledge
• much in common with SW goals of knowledge
  representation and processing but with real-time
  and uncertainty concerns

4
AFRL Research Focus

• US AFRL supported effort to formalize and
  automate the identification and monitoring of
  relevant relations in evolving situations
• Phase I Formalization of SAW
• Formal definition of situation awareness using
  Speckware and DAML/OWL
• Phase II SAW Assistant (SAWA)
• Prototype system to support the detection and
  monitoring of relevant relations

5
General Methodology

• Working with Subject Matter Experts we first
• develop ontologies for describing domain-specific
  object classes and properties
• develop rules to define complex relations that
  are grounded in observable data annotated by the
  ontologies
• We then
• populate an inference engine with ontologies and
  domain rules
• establish an input stream of events describing
  object observations annotated using the domain
  ontologies
• use the inference engine to process the event
  stream and detect evolution of higher-order
  relations

6
Ontologies and Rules

• Need ways to represent domain knowledge
  concerning
• Situation Objects, their Attributes and their
  inter-Relations
• OWL provides a solid basis for these needs
• Formal semantics facilitates reasoning with
  generic reasoners (e.g. Jess with OWL axioms)
• Reuse of existing of tools (parsers, consistency
  checkers, etc)
• Main drawback limited representational power
• SWRL used to represent more complex relations
• Permits representation of more complex
  relationships
• e.g. there are two Units U1 and U2 in-region R
  AND U1 is a member-of a Force F1 AND Unit U2 is a
  member-of Force F2 AND F1 is not equal to F2
• Has advantage of formal semantics defined as an
  extension to the semantics of OWL DL
• Easily converted into Jess rules using XSLT

7
SAW Core Ontology
8
SAWA Architecture
SAWA
Knowledge Management
Runtime System
ConsVISor
GUI
SMC
RMA
Consistency Checker
OWL SWRL
domain knowledge
Protégé
goals and queries
RuleVISor
TDB
EMC
SWRL Editor
Ontology Editor
event annotations
9
SAWA Knowledge Management
10
SAWA Runtime System

• SMC Situation Management Comp.
• EMC Event Management Comp.
• TDB Triples Database
• RMA Relation Monitoring Agent
• GUI Graphical User Interface
• Java Components (RMI)
• RMA and TDB
• Based on Jess inference engine
• Store RDF triples in working memory
• Includes OWL axioms for inferring implicit
  triples
• Plus procedural attachments for SWRL built-ins
• TDB supports OQL and What-if reasoning

SMC
TDB
RMA
GUI
EMC
events
ontologies and rules
11
SAWA Runtime GUI
12
Supply Line Scenario

• Simple Scenario hasSupplyLine
• defines a unit to be in supply if a series of
  roads can be traced from the unit to a supply
  station through regions under friendly control

13
Simple Supply Logistics Ontology
14
hasSupplyLine Rule Set
ltrule rlab"has Supply Line"gt ltbodygt
lthslinRegion sub"?unit"
data"?region"/gt lthslisSuppliable
sub"?region" data"true"/gt lt/bodygt
ltheadgt lthslhasSupplyLine sub"?unit"
data"true"/gt lt/headgt lt/rulegt ltrule
rlab"isSuppliable"gt ltbodygt
lthslhasSupplyStation sub"?region"
data"true"/gt lthslunderFriendlyControl
sub"?region" data"true"/gt lt/bodygt
ltheadgt lthslisSuppliable sub"?region"
data"true"/gt lt/headgt lt/rulegt ltrule
rlab"isSuppliable2"gt ltbodygt
lthslconnects sub"?road"
data"?region1"/gt lthslconnects
sub"?road" data"?region2"/gt
ltswrlbnotEqual arg1"?region1"
arg2"?region2"/gt lthslisPassable
sub"?road" data"true"/gt
lthslisSuppliable sub"?region2" data"true"/gt
lt/bodygt ltheadgt lthslisSuppliable
sub"?region1" data"true"/gt lt/headgt
lt/rulegt
ltrule rlab"underFriendlyControl"gt ltbodygt
lthslinRegion sub"?unit"
data"?region"/gt lthslmemberOf
sub"?unit" data"?force"/gt
lthslFriendlyForce ind"?force"/gt lt/bodygt
ltheadgt lthslunderFriendlyControl
sub"?region" data"true"/gt lt/headgt
lt/rulegt ltrule rlab"isPassable"gt ltbodygt
lthslconnects sub"?road"
data"?regionA"/gt lthslconnects
sub"?road" data"?regionB"/gt
ltswrlbnotEqual arg1"?regionA"
arg2"?regionB"/gt lthslunderFriendlyControl
sub"?regionA" data"?force1"/gt
lthslunderFriendlyControl sub"?regionB"
data"?force2"/gt lt/bodygt ltheadgt
lthslisPassable sub"?road" data"true"/gt
lt/headgt lt/rulegt ltrule rlab"hasSupplyStation"gt
ltbodygt lthslinRegion sub"?X"
data"?region"/gt lthslSupplyStation
ind"?X"/gt lt/bodygt ltheadgt
lthslhasSupplyStation sub"?region"
data"true"/gt lt/headgt lt/rulegt
15
SAWA Runtime GUI
16
Lessons Learned

• OWL is very useful for base ontology
  representation
• Pros tools, formal semantics, extensible,
  triples-based
• Cons triples-based (binary predicates), lack of
  complex implication
• Rules on top of OWL is an effective way to
  utilize the benefits of OWL while overcoming some
  of its limitations
• Limitations of SWRL
• restricted to binary predicates - not always
  natural and work around is cumbersome
• built-ins defined as axioms but are needed as
  functions
• need gensym() and assert() to generate and assert
  new entities
• need for negation as failure
• syntax not intended for human processing
• SWRL is too low level of a language for knowledge
  engineering even with a graphical editor
• need higher-level language(s) that can be
  automatically translated to low-level triples
  representation

17
www.vistology.com