Intro to Expert Systems Paula Matuszek CSC 8750, Fall, 2004 - PowerPoint PPT Presentation


PPT – Intro to Expert Systems Paula Matuszek CSC 8750, Fall, 2004 PowerPoint presentation | free to download - id: 1b0478-OTNhM


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation

Intro to Expert Systems Paula Matuszek CSC 8750, Fall, 2004


Cognitive science research in nature of expertise. Expertise = knowledge. Dendral. Mycin. Expert Systems, Paula Matuszek. 4. 10/24/09. Elements of an Expert System ... – PowerPoint PPT presentation

Number of Views:67
Avg rating:3.0/5.0
Slides: 14
Provided by: paulaam


Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: Intro to Expert Systems Paula Matuszek CSC 8750, Fall, 2004

Intro to Expert SystemsPaula MatuszekCSC 8750,
Fall, 2004
What is an Expert System?
  • A Branch of Artificial Intelligence
  • An expert system is
  • An intelligent computer program
  • That uses knowledge
  • And automated reasoning or inference
  • To solve difficult problems
  • By emulating human expertise
  • In a specialized or delimited area
  • Overlaps with knowledge-based systems
  • May provide solutions, take actions, give advice

How did we get here?
  • Early AI programs
  • Powerful symbolic programming
  • Applied to small amounts of data
  • Second stage heuristics
  • Hardware advances more memory available
  • Cognitive science research in nature of expertise
  • Expertise knowledge
  • Dendral
  • Mycin

Elements of an Expert System
  • Knowledge Base
  • Inference
  • Fact/Database
  • Optional but common components
  • User interface
  • Explanation facility
  • Automated Knowledge acquisition facility
  • Other bells and whistles
  • An expert system shell is a system for creating
    expert systems, typically containing all of the
    above except the KB

Knowledge Base (KB)
  • Information derived from the human expert rules
    of thumb, in-depth knowledge about overall
  • Knowledge about a field represented in an
    organized, declarative fashion
  • Each application of an expert system has a KB
  • Process of creating KB is knowledge engineering
  • KE includes
  • Decisions about What information to represent
  • Decisions about How to represent it
  • Encoding detailed knowledge
  • Roles in KE include
  • Expert system specialist (knowledge engineer)
  • Subject matter specialist (domain expert)

Inference Engine
  • Reasoning tool of the expert system.
  • Inference engine
  • Applies the contents of the knowledge base
  • To the information in a fact base
  • To reach a decision.
  • Inference typically proceeds in one of two ways
  • Start with known facts and work forward (forward
  • Start with a goal and work backward (backward
  • Form of inference is related to form of knowledge

Fact Base
  • Information to this specific case or run of the
    expert system.
  • Also known as database, blackboard, instances
  • Typically empty at the beginning of a case.
  • Gradually instantiated as a case is processed
  • May include information such as
  • Details about case
  • Catalog/price information
  • Configuration or location information
  • Intermediate or temporary information used by
    inference mechanism
  • Often partially instantiated automatically from a

Typical Expert Systems Applications
  • Maintenance and diagnostics
  • Configuration and design
  • Advising
  • Instruction
  • Interpretation
  • Monitoring and control, often real-time
  • Planning

Some Advantages of Expert Systems
  • Availability
  • Available 24/7
  • Available in inaccessible or dangerous locations
  • Consistency and reliability
  • Speed of response
  • Cost
  • Replace human experts with lower-level personnel
  • Replace human experts with machine contacts
  • Permanence (capture knowledge)
  • Easy record keeping, links to dbs, explanations
    of decisions, other uses for knowledge once

What does building an expert system involve?
  • Scope the problem domain
  • Choose appropriate knowledge representation for
    the problem
  • Choose appropriate inference for KR and problem
  • Implement inference or choose tool/shell
  • Create knowledge base
  • Iterate test and modify
  • Add bells and whistles

Characteristics of an Expert System
  • High performance as good as human expert
  • Adequate response time
  • Reliability doesnt crash, fails softly
  • Understandable
  • Builds confidence in decisions
  • Verifies the knowledge
  • Improves knowledge of users
  • Flexible

Good Problems for Expert Systems
  • Theoretical issues
  • Domain is well-defined and delimited
  • Experts exist, and expertise can be taught
  • Problem solving does not have heavy dependency on
    common sense
  • Information needed can be input readily
  • Typically, heuristic domains with a lot of
  • Practical issues
  • Enough use to justify cost of building and
  • Experts are co-operative
  • Experts are rare, inaccessible, expensive,
  • Other more appropriate solutions dont exist

Red Tape and Process
  • Information will be posted at http//
  • Overview and text
  • Syllabus
  • Requirements
  • Academic Integrity
  • Version 6.05 on CD
  • Version 6.2 can be downloaded from download site