Multidimensional Ontological Representation to Enable Knowledge Creation and Optimization within a K

1
Multi-dimensional Ontological Representation to
Enable Knowledge Creation and Optimization within
a Knowledge Management System (KMS) Framework 14
February 2003 Riki Y. Morikawa (PhD
Pre-Candidate) Advisor Dr. Larry
Kerschberg George Mason University School of
Information Technology and Engineering
2
High Level Problem Statement(s)

• Typical Web searches result in the return of
  numerous irrelevant links to information. This
  results in inefficient use of network resources
  and user time. In addition, poor search results
  are not conducive to either explicit or tacit
  knowledge creation.
• Develop a construct that takes into consideration
  the multi-dimensional aspects of a KMS
  architecture
• Temporal, Spatial, Categorical (Ontology),
  Task-Based Vs. Non-Task-Based, Relevancy, Event
  Probability
• Similar to the difficulties faced by IT
  organizations, it is difficult to justify the
  high costs of developing and implementing a
  Knowledge Management System.
• Develop a method that can demonstrate logical and
  physical network efficiency gains that result
  from the adoption of a KMS
• Assumption By developing and optimizing a
  Knowledge Management System, you optimize the
  underlying physical and logical network.

3
What is Knowledge?

• Data is raw numbers and facts, Information is
  processed data, and Knowledge is information
  made actionable (Maglitta, 1996)
• INFORMATION ACTION KNOWLEDGE
• Early information technologies were designed to
  disseminate vast amounts of information to
  managers (MIS)
• Many of these systems evolved to provide decision
  making tools (DSS)
• Knowledge Management Systems (KMS) focus on the
  creation, gathering, organizing, and
  disseminating an organizations knowledge as
  opposed to information or data

4
Ontology
Described by DAML (DARPA Agent Markup Language)
OIL (Ontology Inference Layer) provides a rich
set of constructs with which to create ontologies
and to markup information so that it is machine
readable and understandable.
EXAMPLE Three government organizations have
specific objectives. Although objectives differ,
information and knowledge can overlap, and
therefore efficiencies can be gained and
knowledge creation enhanced.
FBI LAW ENFIRCEMENT
CIA INTELLIGENCE
DRUGS FRAUD KIDNAPPING COUNTER- TERRORISM
COUNTER- NARCOTICS PROLIFERATION FOREIGN
INTEL TERRORISM
COUNTER- DEFENSIVE READINESS TERRORISM SPACE R
D
DoD DEFENSE
5
Ontology - example
e.g. Foreign Narcotics Smuggling

• Ontology overlap exists in several categories
• Subject or Objective Law Vs. Intel.
• Temporal chronology of events where one event
  is the cause of another
• Spatial events occur in regions of interest, or
  regions that are linked together by the
  suspicious activities

COUNTERTERRORISM
FBI LAW ENFIRCEMENT
CIA INTELLIGENCE
SUBJECT
DoD DEFENSE
e.g. Satellite Development
e.g. Threats against US Military Facilities
TEMPORAL (DATE/TIME)
SPATIAL (REGION)
6
Beginnings of a KMS Architecture contd

• An ontology gives information meaning and
  context, however, probabilities of occurrence and
  levels of relevance add to the accuracy of the
  information and decision making process.
• A string of events can make an issue/data item
  relevant (e.g. 911)
• Markov Chains
• FTA, FMEA, etc. (e.g. Firestone Tire Failure)
• Relevance assigned by analyst or intelligent
  software

r

• Pi,j ? Pi,k Pk,j

k 1
high
Increasing Threat
probability
low
low
high
relevance
7
Knowledge Representation Object (KDM Model)
ltOIDgt XXXXX lt/OIDgt ltSUPERTYPESgt OBJECT TYPE
lt/SUPERTYPEgt ltSUBTYPESgt OBJECT TYPE
lt/SUBTYPEgt ltATTRIBUTESgt ltDESCRIPTIONgt XXX
lt/DESCRIPTIONgt ltTEMPORALgt XXX lt/TEMPORALgt
ltSPATIALgt XXX lt/SPATIALgt ltCATEGORICALgt XXX
lt/CATEGORICALgt lt/ATTRIBUTESgt ltALIASgt XXX
lt/ALIASgt ltEVENT CHAINgt ltPREDECESSORgt ltP1gt
0.XXX lt/P1gt lt/PREDECESSORgt ltSUCCESSORgt
ltP2gt 0.XXX lt/P2gt lt/SUCCESSORgt ltRELEVANCYgt XXX
lt/RELEVANCYgt
HIERARCHY
TEMPORAL, SPATIAL, CATEGORICAL
EVENT CHAIN
RELEVANCY MEASURE
KDM Meta-Object Model (Kerschberg)
8
Investigations
Intelligent Thesaurus
Submit Data, Make Queries
XML DB

• Best Methods (doc, epath, content) for
  determining Event Chain Probabilities
• Markov Model, FTA or FMEA for well understood
  models
• Expand BITCUBE Concept to include
  Multi-Dimensional Indexing
• Investigate Effect of KMS and Multi-Dimensional
  Indexing on the Underlying Physical and Logical
  Infrastructure

9
Back-up Slides
10
Ontology

• Ontology provides meaning to information that can
  be actionable however, probabilities of
  occurrence and levels of relevance add to the
  accuracy of the information that benefit the
  decision making process.
• A string of events can make an issue/data item
  relevant
• A string of events within an ontology can have
  associated probabilities of occurrence and levels
  of relevance. From these, a factor of confidence
  based upon observation, and explicit/implicit
  knowledge can be determined.
• Each event is information. This information
  contributes to what we know about certain
  subjects thus adding to the body of explicit
  knowledge. Experienced knowledge workers within
  organizations provide the tacit knowledge needed
  to draw conclusions and assign probabilities.

11
KM Architecture Stack
Tacit or Implicit Knowledge Mental models and
experiences of individuals (Bourdreau and
Couillrd, 1999).
Implicit Knowledge
Explicit Knowledge Formal models, rules and
procedures.
Explicit Knowledge
Ontology Specification of a conceptualization.
DAMLOIL are used to define an ontology.
Ontology
SOAP
SOAP (Simple Object Access Protocol) Protocol
based upon XML designed to connect computers
regardless of their operating system, programming
languages, or object model.
RDF
XML
HTTP
RDF (Resource Description Framework) Provides
support for semantic descriptions of data
however, RDF is weak in their support of
structural and data typing constraints. (see XSD)
TCP
IP
Physical
XML (eXtensible Markup Language) Similar to
HTML, however allows user defined tags. This
enables communicating parties to define common
terms of reference.
12
HTML Vs. XML
lthtmlgtltheadgtlttitlegtCounterterrorist Incident
Reportlt/titlegtlt/headgt ltpgtDate JAN
2000ltBRgt Place Kuala Lumpur, MalaysialtBRgt Individ
uals Khalid Almidhar, Nawaf Alhamzi ltpgtRemarks
meeting captured on Surveillance tapelt/pgt lt/htmlgt
Fixed tags that do not convey meaning to enclosed
data
html

• More than one namespace may appear on an XML
  document

Namespaces help to define ontology
Tags are selected by communicating parties. Tags
are given meaning.
lt?xml version1.0?gtlttitlegtCounterterrorist
Incident Reportlt/titlegt ltCIR xmlnsCThttp//cia.
ctc.gov/ct_schemagt ltdategtJAN 2000lt/dategt ltplacegt
Kuala Lumpur, Malaysialt/placegt ltindividualgtKhalid
Almidharlt/individualgt ltindividualgtNawaf
Alhamzilt/individualgt ltremarksgtmeeting captured on
surveillance tapelt/remarksgt lt/CIRgt
xml
13
(No Transcript)
14
(No Transcript)
15
Lines of Code for XML Parser Ex.
Open "d/example2.txt" For Input As 1 a
Input(LOF(1), 1) RichTextBox1 a strData
a filelength LOF(1) Close 1 For y 1 To 20 x
iEnd iBegin InStr(x, strData, "lt") iEnd
InStr(iBegin 1, strData, "gt") e Mid(strData,
iBegin 1, iEnd - iBegin - 1) If e "\xml"
Then GoTo 10 MSFlexGrid1.TextMatrix(z, 0)
e iBegin InStr(iEnd, strData, "gt") iEnd
InStr(iBegin 1, strData, "lt") f Mid(strData,
iBegin 1, iEnd - iBegin - 1) MSFlexGrid1.TextMat
rix(z, 1) f iBegin InStr(iEnd, strData,
"lt") iEnd InStr(iBegin 1, strData, "gt") g
Mid(strData, iBegin 1, iEnd - iBegin -
1) MSFlexGrid1.TextMatrix(z, 2) g z z
1 Next y
16
Resource Description Framework (RDF)

• Applies description/meaning to information
  resources
• Goes a step beyond XML in providing semantics to
  data (meta data)
• Provides for the concept of triples (subject,
  predicate, object)
• Viewed as a relationship between two resources
• Makes use of namespaces and Dublin Core (1995,
  OH)

author
book
name
17
Beginnings of a KMS Architecture
The Semantic Web layer cake as presented by Tim
Berners-Lee
KM Architecture example- Kerschberg, 2001
C
DS
.05
1
0.5
C
C
WS
WS
AS
AS
DS
1
0.95
0.5
1
1
1
Server utilization example Browser search
probabilities (Menasce, INFT818, 2001)
18
The Semantic WEB
The Semantic Web is an extension of the current
web in which information is given well-defined
meaning, better enabling computers and people to
work in cooperation. - Tim Berners-Lee, James
Handler, Ora Lassila

• The Semantic Web will contain contents and
  services that are understood by computers and
  software agents, and will be based upon XML, RDF,
  and WSDL.
• Will most likely consist of several specialized
  ontologies connected by pointers and data
  populated by users.
• Provides an infrastructure that enables the use
  of remote, mobile software agents.
• While XML Schemas can provide the strong data
  typing, RDF can provide the meaning.

19
The Semantic WEB

• XML Linking Language (XLink) - describes a
  standard way to link XML data and documents
  together.
• expresses information on the nature and behavior
  of the link, support multiple destinations, allow
  link authors to define endpoints and traversal
  rules, backward compatible
• link definition defines the relationship
  between the items to the linked
• participating resources items connected
  together
• traversal rules or arcs direction of
  traversal between pairs
• XML Path Language (XPath) enables hierarchical
  nodes sets to be created that consist of a root,
  elements, attributes, namespaces, processing
  instructions, and comments.
• Namespaces provides meaning to attributes and
  elements within an XML document. Schemas can
  have their own namespaces in which all
  declarations are unique and have their own
  meanings. This helps to deconflict attributes
  that have different meanings.

20
How to describe an Ontology

• Ontology establishes a joint terminology
  between members of a community of interest.
  Members can be human or automated agents.
• XML similar to HTML, however, it leaves the
  interpretation of the data to the end application
• Verbose by design (much larger than binary files)
• Typically uses HTTP as a communications protocol
  (HTTP/1.1 enables fast data compression)
• Development started in 1996 and has been a W3C
  Recommendation since February 1998
• Modular namespaces such as XSL and RDF define
  document formats that are re-usable
• Namespaces are a collection of names, identified
  by a URI, which are used in XML documents as
  element types or attribute names
• Basis for RDF and the Semantic Web
• RDF supports resource description and metadata
  applications