Knowledge Modelling: Foundations, Techniques and Applications

1
Knowledge ModellingFoundations, Techniques and
Applications

• Enrico MottaKnowledge Media InstituteThe Open
  UniversityUnited Kingdom

2
Basic KBS Architecture
Inference Engine
User Interface
Domain Knowledge Base
3
First Generation KBS Architecture
Inference Engine
User Interface
Rule-based Backward-chaining
Domain Knowledge Base
Set of Domain rules
4
Problems

• Focus on implementation-level aspects (backward
  chaining) rather than knowledge-level
  functionalities (medical diagnosis)
• Poor explanation capabilities
• Difficult to assess competence
• Low-level reuse support
• Rules tend to be application specific

5
Heuristic Classification Model
Clancey, AI Journal, 27, 1985
Data Abstractions
Solutions Abstractions
Heuristic Match
Abstraction
Refinement
Solutions
Data
6
HC in Medical Diagnosis
Gram-negative Infection
Data Abstractions
Solutions Abstractions
Heuristic Match
Immunosuppressed
Refinement
Abstraction
Solutions
Data
E-coli Infection
Low white blood count
7
HC in Book Selection
Intelligent Book
Data Abstractions
Solutions Abstractions
Heuristic Match
Educated Person Stereotype
Refinement
Abstraction
Solutions
Data
Watches no TV
Anna Karenina
8
So What? (Competence vs Performance)

• Knowledge-level analysis shows what system
  actually does, not how it does it
• The interesting aspect about Mycin is its
  classification behaviour, not its depth-first
  control regime
• Separation of competence from performance (or
  specification from implementation)
• Important for both analysis and design of
  knowledge-intensive systems

9
So What? (Levels of system analysis)

• There exist different levels at which a system
  can be described
• knowledge-level (tasks and problem solving
  methods)
• Symbol-level (backward-chaining)
• Sub-symbol level (registers)
• Shift in the level of analysis
• Wrong question Can a problem be solved by means
  of a rule-based system?
• Right questions What type of knowledge-intensive
  task are we tackling? What are the appropriate
  problem solving methods?

10
So What? (Reuse)

• Knowledge-level analysis uncovers generic
  reasoning patterns in problem solving agents
• E.g., heuristic classification
• Shift from rule-based reuse to knowledge-level
  reuse
• Focus on high-level reusable task models and
  reasoning patterns
• Classes of tasks
• Design, diagnosis, classification, etc.
• Problem solving methods
• Design methods, classification methods, etc.

11
So What? (Research Development)

• Model-based knowledge acquisition
• From acquiring rules to instantiating task models
• Robust KBS development by reuse
• KBS as a structured development process
• Robustness and economy
• Importance of libraries
• KBS development not necessarily an art!
• Towards a practical theory of knowledge-based
  systems
• What are the classes of tasks/problem solving
  methods?
• How do we identify/model them?
• When are methods appropriate?

12
Knowledge-level Architecturesfor Sharing and
Reuse

• Application of the modelling paradigm to the
  specification and use of libraries of reusable
  components for knowledge systems

13
Modelling Frameworks (1)

• A modelling framework identifies the generic
  types of knowledge which occur in knowledge
  systems, thus providing a generic epistemological
  organization for knowledge systems
• Several exist
• KADS/Common KADS - Un.of Amsterdam
• Components of Expertise - Steels
• Generic Tasks - Chandrasekaran
• Role-limiting Methods - McDermott
• Protégé - Musen, Stanford
• TMDA - Motta
• UPML - Fensel Motta

14
Modelling Frameworks (2)

• Much in common
• Emphasis on reusable models
• Typology of generic tasks
• Constructivist paradigm
• Some differences
• Different degrees of coupling between
  domain-specific and domain-independent knowledge
• Different degrees of flexibility
• Different typologies of knowledge categories

15
A Constructive Approach...

• Lets define our own framework...

16
Generic Tasks

• Informal definition
• A generic class of applications - e.g., planning,
  design, diagnosis, scheduling, etc..
• More precise definition
• A knowledge-level, application-independent
  description of the goal to be attained by a
  problem solver.
• Several typologies exist
• e.g., Breuker, 1994
• Viewpoints over applications
• No natural categories
• Different viewpoints can be imposed on a
  particular application

17
Example Parametric Design

• Generic Task Parametric Design
• Inputs Parameters, Constraints, Requirements,
  Cost-Function, Preferences
• Output Design-Model
• Goal To produce a complete and consistent
  design model, which satisfies the given
  requirements
• Preconditions At least one requirement and
  one parameter are provided

18
Example Classification

• Generic Task Classification Inputs Candidate-cla
  sses Observables
• Output Best-Matching-Classes
• Preconditions At least one candidate class
  exists
• Goal To find the class that best explains
  the observables

19
Generic Component 2 Reusable PSMs

• A domain-independent, knowledge-level
  specification of problem solving behaviour, which
  can be used to solve a class of tasks.
• PSM specifications may be partial
• PSM can be task-specific
• E.g., heuristic classification
• PSM can be task-independent
• E.g., search methods, such as hill-climbing, A,
  etc.....

20
Functional Specification of a PSM

• Problem solving method search
• ontology
• import
• state-space-terminology
• competence
• roles
• input input State
• output output State
• preconditions
• input ? 0
• postconditions
• solution_state (output)
• assumptions
• ?s . solution_state (?s) successor
  (input, ?s)

21
Operational Description

• Begin
• states one x. initialize (input input)
• repeat
• state one x . select _state (states states)
• if solution_state (state)
• then return state
• else
• succ_states one x. derive_successor_states
  (state state)
• states one x. update_state_space (input1
  states input2 succ_states)
• end if
• end repeat
• end

22
Task-Method Structures
Problem Type
Primitive PSM
23
Multi-Functional Domain Models

• Domain-specific models, which are not committed
  to a specific PSM or task.
• Examples
• A database of cars
• The CYC knowledge base, etc..

24
Picture so far..
Problem Solving Method
Generic Task
Simple Classifier
Classification
Multi-Functional Domain
Lunar rocks
25
Issue

• How to link different reusable components?

Application Model
Classification
Simple Classifier
Problem Solving Method
Generic Task
Multi-Functional Domain
Lunar rocks
26
Solution Mappings

• Mappings model explicitly the relationship
  between different components in an application
  model

Application Model
Classification
Simple Classifier
Task-PSMMapping
Problem Solving Method
Generic Task
PSM-DomainMapping
Task-DomainMapping
Multi-Functional Domain
Lunar rocks
27
Example

• Scenario Office Allocation Application
• Generic Task Parametric Design
• Domain KB about employees and offices

Task Level
Parameter
Design Model
Domain Level
Pairs ltEmployee, Roomgt
Employee
28
Application-specific knowledge

• Mappings are an example of application-specific
  knowledge. Are there others?

Yes Application-specific heuristic problem
solving knowledge
29
Elevator Design Example

• A configuration designer only considers two
  positions for the counterweight
• Half way between platform and U-bracket
• A position such that the distance between the
  counterweight and the platform is at least 0.75
  inches

30
Complete Picture
Application Model
Generic Task
Problem Solving Method
Mapping Knowledge
Application-specific Problem-Solving Knowledge
Application Configuration
Multi-Functional Domain
31
Even More Complete Picture
Application Model
Generic Task
Problem Solving Method
Task Ontology
Method Ontology
Mapping Knowledge
Application-specific Problem-Solving Knowledge
Ontology
Mapping Ontology
Application Configuration
Multi-Functional Domain
Domain Ontology
32
(No Transcript)