Notes
Outline
DAML Experiment
Semantic Operational Net Assessment Tool (SONAT)
2002 Status Report
John Flynn
[email protected]
Mike Dean
[email protected]
DAML Experiment Goals
Effects Based Operations and
Operational Net Assessment
Slide 4
Slide 5
Slide 6
Slide 7
Slide 8
Slide 9
Slide 10
SONAT Architecture
SONAT GUI
DAML.NET API
Object Viewer
Agent Infrastructure
Slide 16
ENP Agent
Direct Effects
Slide 19
Indirect Effects Agent
Computes indirect effects based on direct effects specified in the ENP
Necessary because DAML+OIL doesn't support transitive n-ary relations
Current implementation uses cwm rules
Very concise definition
Want combination of --filter and --think
Resulting statements postprocessed to tag with agent
Could alternatively use SweetJess
Possible next step:  use JTP rules and tag each resulting statement with a DAML+OIL representation of its justification
Slide 21
MCDA Agent
Slide 23
Matching Targets’ Features and Units’ Capabilities
Targets are described by a set of features
ID: geonames?f9086465
Type: Bridge
Features:
Fixed
Linear
Point
Soft
Lat 37.1330556
Long 73.9947222
Units can attack targets with specified features
Unit Name: 13th MEU (SOC)
Features:
Area
Fixed
Linear
Personnel
Point
Soft
Lat 24.9745
Long 65.53
JBI BDA Agent
Metrics Agent
Collects metrics on ontologies and content used in DAML Experiment
Lessons Learned
DAML allows data to be linked rather than copied
Most data sources have fairly simple ontologies
Often preferable to add properties to existing instances (e.g., countries) rather than creating new instances
Most SONAT data sources and agents can be viewed as services providing properties for a class of instances (information services)
DAML worked well as agent description and message format
Use of C# and .NET required significant effort but resulted in a set of tools to help future developers
Browser-only implementation is feasible
Knowledge can be effectively distributed
DAML Experiment 2003
More DAML technologies
Incremental integration
More agents
Cooperating-agents
Agent infrastructure
DAML Services
Query
Inference
Semantic integration
BACKUP
Slide 30
Matching Units and Targets
Problem:
Through SONAT the officer can establish the targets to hit in a given operation
The officer can also specify which units are available for the operation
Units/Targets matcher associates a unit to a target
Matching Process
Use Jena to load ontologies
Compile Ontologies into Jess
    resulting in a KB of about 2500 axioms
Add inheritance rules to Jess model
    inference adds  about 300 additional axioms
Perform inference on Jess model to find units matching targets
Tie-breaking rule selects closest unit to target
Translation from Jena to Jess
Translation of triples:
A triple <subject predicate object>
is transformed into a predicate  predicate(subject,object)
2 inference rules for inheritance are added:
Transitivity of subclassing
subclass(X,Y) Ù subclass(Y,Z) Þ subclass(X,Z)
Distribution of type
type(X,Y) Ù subclass(Y,Z) Þ type(X,Z)
Ouput
Output in DAML reports the selected unit and the units that also match the target
<resultOnt:selectedUnit>
    <resultOnt:UnitResult>
       <resultOnt:unit rdf:resource="http://www.daml.org/…/unit#FFJFD0"/>
        <resultOnt:distance>2.281111211937926      </resultOnt:distance>
    </resultOnt:UnitResult>
 </resultOnt:selectedUnit>
  <resultOnt:matchedUnit>
    <resultOnt:UnitResult>
      <resultOnt:unit rdf:resource="http://www.daml.org/…/unit#N09600"/>
      <resultOnt:distance>13.839447605644123      </resultOnt:distance>
    </resultOnt:UnitResult>
  </resultOnt:matchedUnit>
"The following slides describe the..."
The following slides describe the matchmaker
CMU’s DAML-S Matchmaker
Uses DAML-S Profiles to describe advertisements of services and requests of services
Match based on subsumption relations:
Outputs Advertisement É Outputs Request
Inputs Requests É Inputs Advertisement
CMU’s DAML-S Matchmaker
Uses UDDI Registry for storage of advertisements
Extends UDDI Registry adding matching of capabilities that is not available in UDDI.