KNOWLEDGE BASED SYSTEM

Introduction
Peter Obiria
 MOTIVATION

 Utilization of computers to deal with knowledge

 quantity of knowledge available increases rapidly
 relieve humans from tedious tasks
 Computers have special requirements for dealing with
knowledge
 acquisition, representation, reasoning
 Some knowledge-related tasks can be solved better
by computers than by humans
 cheaper, faster, easily accessible, reliable
 OBJECTIVES
 To know and comprehend the main principles,
components, and application areas for expert systems
 To understand the structure of expert systems
 knowledge base, inference engine
 To be familiar with frequently used methods for
knowledge representation in computtasksers
 To evaluate the suitability of computers for specific
tasks
 application of methods to scenarios or
TERMINOLOGY

 Data
 Information
 Knowledge
 DATA PYRAMID AND COMPUTER-BASED SYSTEMS

IS

Strategy makers apply morals, principles,

and experience to generate policies WBS Wisdom (experience)

Higher management generates knowledge

by synthesizing information KBS Knowledge (synthesis)

Middle management uses reports/info.

generated though analysis and acts accordingly
DSS, MIS Information (analysis)

Basic transactions by operational TPS Data (processing of raw observations )

staff using data processing

Volume Sophistication and

complexity

5
 DATA PYRAMID AND COMPUTER BASED SYSTEMS

Heuristics and
models Wisdom

Rules Knowledge
Novelty

Information

Concepts

Experience
Data

Raw data through fact finding Understanding

Researching Absorbing Doing Interacting Reflecting

6 Figure 1.6: Convergence from data to intelligence

 WHAT IS AN EXPERT SYSTEM
(ES)?
 Relies on internally represented knowledge to
perform tasks
 Utilizes reasoning methods to derive appropriate
new knowledge
 Usually restricted to a specific problem domain
 some systems try to capture common-sense
knowledge
 General Problem Solver (Newell, Shaw, Simon)
 Cyc (Lenat)
 DEFINITIONS “EXPERT SYSTEM”
 a computer system that emulates the decision-
making ability of a human expert in a restricted
domain [Giarratano & Riley 1998]
 Edward Feigenbaum
 “An intelligent computer program that uses
knowledge and inference procedures to solve
problems that are difficult enough to require
significant human expertise for their solutions.”
[Giarratano & Riley 1998]
 the term knowledge-based system is often used
synonymously
MAIN COMPONENTS OF AN ES
User
Expertise
Knowledge Base

User Interface
Facts / Information

Inference Engine
Expertise
Developer
COMPONENTS OF KBS

Knowledge base is a repository of

domain knowledge and metaknowledge. Enriches the
Inference engine is a software program system with
that infers the knowledge available in self-learning
the knowledge base. capabilities

Knowledge Inference
Explanation base engine
Self-
and
learning
reasoning
User interface
Provides
explanation and
reasoning Friendly
facilities interface to
users working
Figure 1.10: General structure of KBS in their native
language

10
 MAIN ES COMPONENTS
 knowledge base
 contains essential information about the problem domain
 often represented as facts and rules
 inference engine
 mechanism to derive new knowledge from the knowledge
base and the information provided by the user
 often based on the use of rules
 user interface
 interaction with end users
 development and maintenance of the knowledge base
GENERAL CONCEPTS AND
CHARACTERISTICS OF ES
 knowledge acquisition
 transfer of knowledge from humans to computers
 sometimes knowledge can be acquired directly from the environment
 machine learning
 knowledge representation
 suitable for storing and processing knowledge in computers
 inference
 mechanism that allows the generation of new conclusions from
existing knowledge in a computer
 explanation
 illustrates to the user how and why a particular solution was generated
DEVELOPMENT OF ES
TECHNOLOGY
 Strongly influenced by cognitive science and
mathematics
 the way humans solve problems
 formal foundations, especially logic and inference
 Production rules as representation mechanism
 IF … THEN type rules
 reasonably close to human reasoning
 can be manipulated by computers
 appropriate granularity
 knowledge “chunks” are manageable both for
humans and for computers
[Dieng et al. 1999]
 RULES AND HUMANS

 Rules can be used to formulate a theory of human

information processing (Newell & Simon)
 rules are stored in long-term memory
 temporary knowledge is kept in short-term memory
 sensory input or thinking triggers the activation of rules
 activated rules may trigger further activation
 a cognitive processor combines evidence from
currently active rules
 This model is the basis for the design of many rule-based
systems
 also called production systems
EARLY ES SUCCESS STORIES

 DENDRAL
 identification of chemical constituents
 MYCIN
 diagnosis of illnesses
 PROSPECTOR
 analysis of geological data for minerals
 discovered a mineral deposit worth $100 million
 XCON/R1
 configuration of DEC VAX computer systems
 saved lots of time and millions of dollars
THE KEY TO ES SUCCESS

 convincing ideas
 rules, cognitive models
 practical applications
 medicine, computer technology, …
 separation of knowledge and inference
 expert system shell
 allows the re-use of the “machinery” for
different domains
 concentration on domain knowledge
 general reasoning is too complicated
WHEN (NOT) TO USE ESS
 expert systems are not suitable for all types of
domains and tasks
 conventional algorithms are known and
efficient
 the main challenge is computation, not
knowledge
 knowledge cannot be captured easily
 users may be reluctant to apply an expert
system to a critical task
ES TOOLS

 ES languages
 higher-level languages specifically designed
for knowledge representation and reasoning
 SAIL, KRL, KQML, DAML, OWL (Web Ontology
Language)
 ES shells
 an ES development tool/environment where
the user provides the knowledge base
 CLIPS, JESS, Mycin, Babylon, G2, ...
RELATED DEVELOPMENTS

 Semantic Web
 extension of the World Wide Web
 includes knowledge representation and reasoning
capabilities
 Decision Support Systems
 less emphasis on autonomy
 Data Mining
 extraction of knowledge from large quantities of
data
 Sensemaking
 computer support for quicker, easier understanding
of complex domains or situations
ES ELEMENTS

 knowledge base
 inference engine
 working memory
 agenda
 explanation facility
 knowledge acquisition facility
 user interface
ES STRUCTURE

Knowledge Base
Knowledge
Acquisition
Facility
User Interface

Inference Engine Agenda

Explanation
Facility
Working Memory

Introuction
RULE-BASED ES
 knowledge is encoded as IF … THEN rules
 these rules can also be written as production rules
 the inference engine determines which rule antecedents
are satisfied
 the left-hand side must “match” a fact in the working
memory
 satisfied rules are placed on the agenda
 rules on the agenda can be activated (“fired”)
 an activated rule may generate new facts through its
right-hand side
 the activation of one rule may subsequently cause the
activation of other rules
EXAMPLE RULES
IF … THEN Rules
antecedent
Rule: Red_Light
(left-hand-side)
IF the light is red
THEN stop
Rule: Green_Light consequent
IF the light is green (right-hand-side)
THEN go

Production Rules antecedent (left-hand-side)

the light is red ==> stop
consequent
(right-hand-side)
the light is green ==> go
MYCIN SAMPLE RULE
Human-Readable Format
IF the stain of the organism is gram negative
AND the morphology of the organism is rod
AND the aerobiocity of the organism is gram anaerobic
THEN the there is strongly suggestive evidence (0.8)
that the class of the organism is enterobacteriaceae
MYCIN Format
IF (AND (SAME CNTEXT GRAM GRAMNEG)
(SAME CNTEXT MORPH ROD)
(SAME CNTEXT AIR AEROBIC)
THEN (CONCLUDE CNTEXT CLASS ENTEROBACTERIACEAE
TALLY .8)
[Durkin 94, p. 133]
 INFERENCE ENGINE CYCLE

 describes the execution of rules by the inference engine

 conflict resolution
 select the rule with the highest priority from the agenda
 execution
 perform the actions on the consequent of the selected rule
 remove the rule from the agenda
 match
 update the agenda
 add rules whose antecedents are satisfied to
the agenda
 remove rules with non-satisfied agendas
 the cycle ends when no more rules are on the agenda, or when an
explicit stop command is encountered
 FORWARD AND BACKWARD
CHAINING
 different methods of rule activation
 forward chaining (data-driven)
 reasoning from facts to the conclusion
 as soon as facts are available, they are used to match
antecedents of rules
 a rule can be activated if all parts of the antecedent are
satisfied
 often used for real-time expert systems in monitoring and
control
 examples: CLIPS, OPS5
 backward chaining (query-driven)
 starting from a hypothesis (query), supporting rules and facts
are sought until all parts of the antecedent of the hypothesis
are satisfied
 often used in diagnostic and consultation systems
 examples: EMYCIN
FOUNDATIONS OF EXPERT
SYSTEMS
Rule-Based Expert Systems

Inference Engine Knowledge Base

Pattern
Conflict Facts Rules
Matching
Resolution
Post
Rete
Production
Algorithm Action Rules
Execution
Markov
Algorithm
POST PRODUCTION SYSTEMS
 production rules were used by the logician Emil L. Post in the early
40s in symbolic logic
 Post’s theoretical result
 any system in mathematics or logic can be written as a
production system
 basic principle of production rules
 a set of rules governs the conversion of a set of strings into
another set of strings
 these rules are also known as rewrite rules
 simple syntactic string manipulation
 no understanding or interpretation is required
 also used to define grammars of languages
 e.g. BNF grammars of programming languages
EMIL POST
 20th century mathematician
 worked in logic, formal
languages
 truth tables
 completeness proof of the
propositional calculus as
presented in Principia
Mathematica
 recursion theory
 mathematical model of
computation similar to
the Turing machine
 not related to Emily Post
http://en.wikipedia.org/wiki/Emil_Post
;-)
MARKOV ALGORITHMS
 in the 1950s, A. A. Markov introduced
priorities as a control structure for
production systems
 rules with higher priorities are
applied first
 allows more efficient execution of
production systems
 but still not efficient enough for
expert systems with large sets of
rules
 he is the son of Andrey Markov, who
developed Markov chains
RETE ALGORITHM

 developed by Charles L. Forgy in the late 70s

for CMU’s OPS (Official Production System)
shell
 stores information about the antecedents
in a network
 in every cycle, it only checks for changes
in the networks
 this greatly improves efficiency
RETE NETWORK

http://en.wikipedia.org/wiki/File:Rete.JPG
© 2002-2010 Franz J. Kurfess Introduction
ES ADVANTAGES
 economical
 lower cost per user
 availability
 accessible anytime, almost anywhere
 response time
 often faster than human experts
 reliability
 can be greater than that of human experts
 no distraction, fatigue, emotional involvement, …
 explanation
 reasoning steps that lead to a particular conclusion
 intellectual property
 can’t walk out of the door
ES PROBLEMS
 limited knowledge
 “shallow” knowledge
 no “deep” understanding of the concepts and their relationships
 no “common-sense” knowledge
 no knowledge from possibly relevant related domains
 “closed world”
 the ES knows only what it has been explicitly “told”
 it doesn’t know what it doesn’t know
 mechanical reasoning
 may not have or select the most appropriate method for a
particular problem
 some “easy” problems are computationally very expensive
 lack of trust
 users may not want to leave critical decisions to machines
SUMMARY INTRODUCTION

 expert systems or knowledge based systems are used to

represent and process in a format that is suitable for
computers but still understandable by humans
 If-Then rules are a popular format
 the main components of an expert system are
 knowledge base
 inference engine
 ES can be cheaper, faster, more accessible, and more
reliable than humans
 ES have limited knowledge (especially “common-sense”),
can be difficult and expensive to develop, and users may
not trust them for critical decisions
IMPORTANT CONCEPTS AND
TERMS
 knowledge base
agenda ❖
 backward chaining ❖ knowledge-based system
 common-sense ❖ knowledge representation
knowledge
 conflict resolution
❖ Markov algorithm
 expert system (ES) ❖ matching
 expert system shell ❖ Post production system
 explanation ❖ problem domain
 forward chaining

❖ production rules
inference
 inference mechanism
❖ reasoning
 If-Then rules ❖ RETE algorithm
 knowledge ❖ rule
 knwledge acquisition ❖ working memory