Ontology and Agent based Approach for Knowledge Management

Ontology and Agent based
Approach for Knowledge
Management

Defense of PhD Thesis

Michal Laclav
í
k

Supervisor: Ing. Ladislav Hluch
ý

PhD.

Bratislava, 12th January 2006

2

Outline


Motivation


State of the Art


Objectives


Methodology and Tools


Agent Knowledge Model
–

Models,
Methodology, Library


Experience Management


Applications


Conclusion

Bratislava, 12th January 2006

3

Motivation and State of the Art


MAS is

powerful paradigm for distributed or heterogeneous
systems


MAS need Knowledge Support and Semantics


MAS need Connection with Existing Commercial Standards



Agent Technology Roadmap
: Current MAS Systems
–

lack of
Internal Agent Knowledge Model, lack of interconnection with
semantic web results (knowledge model representations) and
commercial standards



Focus on Agents and Knowledge representation (Ontologies)


Knowledge Management and Experience Management as
application domains

Bratislava, 12th January 2006

4

State of the Art
-

Agents


Agent Definition:

An agent is a computer
system capable of flexible autonomous action in
a dynamic, unpredictable and open environment.

(LUCK 2003)


MAS Standards:

FIPA, MASIF


Related to agent communication, agent platforms


No standards for internal agent knowledge model with
available implementations

Bratislava, 12th January 2006

5

State of the Art
-

Agents


Architectures:


Reactive Architecture


No specification of knowledge model, behavior of agent is based on
implemented responses to environment states


Belief Desire Intention Architecture
–

BDI


Belief
–

represents knowledge model, available some implementations
based on logic programming, not used in FIPA compliant MAS


Behavioral Architecture


FIPA compliant MAS are based on such architecture


No specification of Internal Agent Knowledge model
–

depend on agent
designer and developer


JADE Agent System


Support for ontologies based on FIPA
-
SL (Similar to First Order
Logic)


No Query engine


No Storage


No Inference


Bratislava, 12th January 2006

6

State of the Art
–

Ontologies, Knowledge



Ontologies


Knowledge Representation


OWL
-
DL compatible with Description
Logic


Query and Storage Engines available


RDF, OWL, RDQL based



Application domain


Knowledge Management

(KM) is the
process through which organizations
generate value from their intellectual and
knowledge
-
based assets

(Source: CIO Magazine)


Experience Management
is special
kind of KM
–

based on “lessons learned”


Characters

Data

Information

Knowledge

Actions

Syntax

Semantics

Pragmatics

Reasoning

(Bergman, 2002,

Experience Management)

Bratislava, 12th January 2006

7

Problem Specification

Multi Agent System

Agent 1

Agent 2

Agent 3

Graphical User Interface

External

System

Knowledge

Base

Directory

Facilitator

Knowledge Storage

Querying

XML, XML
-
RPC, SOAP

User requests

Displaying results

FIPA ACL,

KIF, FIPA
-
SL, FIPA
-
RDF

FIPA ACL,

RDF/OWL, RDQL

Knowledge

Model

KM

KM

IIOP,

HTTP,

SMTP

ACL

Bratislava, 12th January 2006

8

State of The Art Conclusion


Focus on
software
, intelligent and
FIPA

compliant
agents


Providing better
semantic infrastructure

(ontologies,
knowledge models
)


Apply basic principles of
software

and
knowledge engineering


Make stronger
connection

between MAS
and existing
commercial technologies

Bratislava, 12th January 2006

9

Thesis Objectives


Design of
Agent Architecture

using

Ontology
based

Knowledge Model


Design of
Software Development Methodology

for creation
of Agents with Ontology based Knowledge Model


Design of
Generic Ontology Model

for
Experience
Management

with extension to different application domains.


Design & Development of
Software Library

for building
Intelligent Agents with Ontology Knowledge Model with
possibility to plug agents to existing commercial technologies


Design and Development of user friendly
Knowledge
Presentation
.


Evaluation of Results

on real pilot operation.

Bratislava, 12th January 2006

10

Used Methods and Methodologies


Knowledge management, system design


Unified Modeling Language
–

UML


CommonKADS, MAScommonKADS


Protégé as Tool for CommonKADS


Formal methods for describing

ontology based models


Description Logic


Graph Ontology representation

Bratislava, 12th January 2006

11

Used Tools and Software


Protégé

Ontology Editor


Support for OWL ontology format


Can be used as modeling tool


JADE

(Java Agent DEvelopment
Framework)


Most developing MAS framework


Compliant with FIPA standards


Jena

–

Semantic Web Framework for Java


Support for OWL
–

best available OWL API


Support for RDQL model querying

Agent Knowledge Model

Objective:

Design of
Agent Architecture

using

Ontology
based

Knowledge Model


Bratislava, 12th January 2006

13

Agent Knowledge Model


Based on Events, Resources, Actions, Actors, Context


Formally Described using Sets, Description Logic
(compatible with OWL
-
DL), Graph Representation


Actor Context updating function/algorithm

(Actor Environment State)


C
A
new

= f
C
(e
a
,C
A
old
)


Resources updating function/algorithm

(result of fulfilled actor goals)


R
A
new

= f
R
(C
A
new
,R
A
old
)

Software Development
Methodology

Objective:

Design of
Software Development
Methodology

for creation of Agents
with Ontology based Knowledge Model



Bratislava, 12th January 2006

15

Development Methodology
(Knowledge Model)


Extending Model with Protégé Editor following
CommonKADS models


Organizational or Environment Model


Task Model


Agent or Actor Model


Includes implementation of algorithms for context and resource
updating


Results


Ontology developed in Protégé which can be exported in
OWL format.


Concrete Algorithms for each actor (often algorithms are
similar or same) which updates actors' context
C
A
new

and
resources
R
A
new
.

Bratislava, 12th January 2006

16

Development Methodology

(System Design)


UML Diagrams for concrete Application Domain


Use Case Diagram


for each agent


agent is taken as system

boundaries


Sequence Diagram


Communication among

agents


Class Diagram


Behaviors are described as methods


Agent Software Library

Objectives:


Design of
Agent Architecture

using

Ontology
based

Knowledge Model


Design & Development of
Software Library

for building
Intelligent Agents with Ontology Knowledge Model with
possibility to plug agents to existing commercial technologies


Bratislava, 12th January 2006

18

Agent Software Library


Support for OWL based Agent Knowledge Model


Support for XML
-
RPC connection to receive event and
send plain XML


Support for agent communication using FIPA ACL with
OWL and RDQL as content languages


Support for Presentation of Ontological Knowledge

(RDF/OWL => plain XML + XSL => HTML)


JADE and Jena Integration


Available on JADE official website

to MAS community

Bratislava, 12th January 2006

19

Agent Library Example

Support for Knowledge and
Experience Management

Objective:

Design of
Generic Ontology Model

for
Experience Management

with extension to
different application domains.

Bratislava, 12th January 2006

21

Extension of Model for
Experience Management


Extended Agent Memory
Model



Workflow Related


WfInstance, WfActivity


ActiveHint


Sub class of resource


Representation of
Experience


Employee

Bratislava, 12th January 2006

22

Algorithms for

EM Extension


Actor (Employee)
Context updating
algorithm


C
A
new

= f
C
(e
a
,C
A
old
)


Resources (Active Hint)
updating algorithm


R
A
new

= f
R
(C
A
new
,R
A
old
)



Bratislava, 12th January 2006

23

Complexity

of algorithms


All depends also on Active Hints
Templates count
–

this does not grow
too fast.


1
st

Case: Constant
–

final count of
context elements (1
-
6)


2
nd

Case: O(n)
–

based on
resource/event count in Memory


3
rd

Case: O(n
2
)
–

based on 2 loops:
events/resources, similar resources


experimental solution because algorithm
used other software e.g. Jena with RDQL
–

it was hard to prove complexity
different way.


Bratislava, 12th January 2006

24

Resource Similarity (3
rd

Case)


Similarity of Ontology
Individuals


Weighted matching of
properties


Similar to CBR
algorithm Weighted
Euclidian Distance


sim({res
1
,res
2
}) = fsim
(



"
{prop
i
}

property
i
.Resource({res
1
})




"
{prop
j
}

property
j
.Resource({res
2
})



{prop
i
}


{prop
j
}

{prop
i
}

DomainClass



DomainClass

Domain



$
{simWeight}

SimilarityWeight



domainClass.SimilarityWeight( DomainClass)

{simWeight}



{weight}

weight .SimilarityWeight( DomainClass)

{simWeight};


S
ij
{weight}/n

)

Presentation of Ontology
based Knowledge

Objective:

Design and Development of user
friendly
Knowledge Presentation
.

Bratislava, 12th January 2006

26

Presentation of Ontology
based Knowledge


Ontology Tree


Browse window


Graph


XSL Transformation


RDF/OWL => Plain XML +
XSL => HTML


Infrastructure to receive
plain XML using XML
-
RPC

Applications

Objective:

Evaluation of Results

on real
pilot operation.

Bratislava, 12th January 2006

28

Pellucid 5FP IST Project


Title:
Platform for Organizationally
Mobile Public Employees


Duration:

Sep 2002
-

Dec 2004


Knowledge Management to support
employees


Workflow based Administration
Processes


To support Employee Mobility in
organization


Agent Architecture based on
autonomous co
-
operating agents

Process Layer

Interaction Layer

Pellucid Architecture

Pellucid Agents

Bratislava, 12th January 2006

29

Pellucid Applications


CDG
, Genoa, Italy

Traffic Light Management


MMBG
, Sanlucar, Spain

Project Management


SADESI
, Seville, Spain

Telephone Incidence
Resolution

Bratislava, 12th January 2006

30

K
-
Wf Grid 6FP IST Project


Work on new

EMBET

architecture


Current state:

User Assistant Agent in K
-
Wf Grid uses model
presented in thesis.


Algorithms presented in chapter 5 were reused with same
improvements and modifications.


Architecture is not Agent based but users of system are
modeled as actors.


Knowledge Model
, its
implementation

and modified

algorithms
presented

in thesis are used


Title:
Knowledge
-
based Workflow System for Grid
Applications


Objectives:
To support workflow construction and
execution with Knowledge


Duration:

Sep 2004
-

Feb 2007

Conclusion and Future Work

Bratislava, 12th January 2006

32

Conclusion (1)


The most significant scientific achievements


Agent knowledge model


Applicable in any discrete environment where actors need to be
modeled


Can be expressed by ontology, sets or description logic


Such model was found useful for:


Simple goal oriented agents


Knowledge Management Solution based on Agents (Pellucid)


Experience Management Solution non agent based (EMBET
System)


Development Methodology


Speed up Knowledge based Agent development for concrete
application domains



Bratislava, 12th January 2006

33

Conclusion (2)


The most significant development achievements


Agent Library


Support for OWL based Agent Knowledge Model


Support for XML
-
RPC connection to receive event and send plain XML


Support for Presentation of Ontological Knowledge (RDF/OWL => plain
XML + XSL => HTML)


Support for agent communication using FIPA ACL with OWL and RDQL
as content languages


JADE and Jena Integration


Available on JADE official website to MAS community

(August
-

December 2005
–

314 downloads
)

Bratislava, 12th January 2006

34

Conclusion (3)


Extension of Work for

Experience Management


Model


Algorithms


Projects


Motivation for solving problems in real
Application


Evaluation of Thesis results

Bratislava, 12th January 2006

35

Future work


RAPORT
APVT project (01/2005
-
12/2007): Research and
development of a knowledge based system to support workflow
management in organizations with administrative processes


model and algorithms will be reused and extended



K
-
Wf Grid

EU 6FP RTD IST project (2004
-
2007)


evaluation on more applications, improvement of context detection



NAZOU

SPVV Project (09/2004
-
11/2007): Tools for acquisition,
organization and maintenance of knowledge in an environment of
heterogeneous information resources


OnTeA semantic annotation
–

not directly related but can be used for
context detection

Thank you !

Thank You for you attention

Many Thanks to my supervisor

Many Thanks to my colleagues

Many Thanks to the Reviewers for their helpful and
constructive comments and for reading my thesis