Expert Systems/AI

1
Expert Systems/AI
2
The four areas of AI

• Natural language processing
• expert systems
• vision and speech systems
• robotics

3
Difficulty of developing a natural-language
database inquiry system

• Human speech is complex
• speech is often dependent upon the context
• questions are often hidden within questions
• human speech is often incomplete
• human speech often requires interpretation

4
Translation Systems

• Translation systems can speed up the translation
  process
• Current systems do require a human translator to
  review computer translations due to
  inconsistencies with language itself

5
Expert System Component Interaction

• System is able to use textual materials to aid
  decision making
• System comprised of facts, rules, and a user
  interface
• Is able to use the knowledge base (facts),
  identify guidelines for using the facts (rules),
  and allow an end user to query the system for
  problem solutions.

TIP State the benefits in termsthat relate
to youraudiences interests, needs,and
preferences.
6
Role of the Inference engine in expert systems

• Inference engine tests the rules in the knowledge
  base and prompts the user for information needed
  to prove a rule true or false.

7
Authoring Systems and their role

• Authoring systems allow content matter experts
  who are inexperienced computer users to develop
  computer-aided instruction systems. Authoring
  systems are shells that require the content
  matter expert to identify facts and rules
• .

TIP Remember, your sense ofconviction and
your involvementwith the content of
thepresentation are criticalto its success.
8
Lets review!

• Knowledge base--the part of an expert system that
  contains specific facts about the expert area and
  any rules the expert system will use to make
  decisions based on those facts
• Inference engine--a program used to apply rules
  to the knowledge base in order to reach decisions
• Expert system shell--a prepackaged expert system
  that lacks only a knowledge base. Can be built
  from scratch or acquired as a package. Shells
  represent the most common and inexpensive way to
  create an expert system.

9
Neural net computing

• Used today by many expert systems
• A technology in which the human brains
  pattern-recognition process is emulated by a
  computer.
• Example fuzzy logic, washing machines that
  determne temperature, other settings, microwaves
  that determine settings based on contents, car
  brakes, buying a home, buying/selling stock.

10
Neural net computing--other applications

• handwriting, speech and image recognition
• recognition of handwritten numerals on checks
• recognition of human faces with accuracy of
  better than 99(faceprinting)
• credit risk assessments--solves credit risk
  assessment problems discerns good/poor credit
  risks
• crime analysis--transaction processing and
  fraudulent patterns
• stock analysis--examines stock market data for
  pattern/determines strategies for buying/selling
  stocks

11
Difference between an expert system and an
intelligent job aid.

• When used as a teaching tool rather than an
  on-the-job coach, an expert system is an
  intelligent job aid.

12
Applications of robotics for office work

• Mail robots--A lawsuit is currently underway in
  which an office worker was attacked by the office
  mail robot who charged her and pinned her
  against a wall in an attempt to crush her
• Computer assisted retrieval systems
• systems that allow physically challenged workers
  to do office work

13
What is meant by

• Isolated word systems--understand human speech in
  terms of distinct words separated by distinct
  pauses
• Continuous speech systems--recognize a predefined
  vocabulary of words spoken at a controlled pace
• Speaker-dependent systems--trained to recognize
  the speech of a single-user
• Speaker-independent systems--recognize speech
  independent of who is doing the speaking.

14
Most potential speech recognition system for
office applications

• Continuous speech, speaker-independent systems
  have the most potential because of their ease of
  use.

15
Can a machine be taught to talk?

• Teaching a machine to respond in audio fashion is
  difficult but not impossible, example VOICE
• Difficulties can be compared to teaching a person
  a foreign language
• Problems with phonetics and mispronunciation are
  common

16
Legality and ethics of expert system ownership

• Problem evolves around who owns the data
• Is the information in the expert system the
  property of the subject matter expert? The
  knowledge engineer? The company?
• Companies are addressing the legal issues prior
  to actual use of the system
• Companies are posing the ethical issues prior to
  system development

17
The inexactness of human language

• To prove the inexactness of human language, take
  the following oral quiz
• How many are a few chairs?
• How far is a few feet?
• How old is an old man?
• How old is an old woman?
• How tall is a tall man?
• How short is a short woman?

18
An expert system at work

• Query Should we issue credit to Mr. Sam for a
  700 purchase?
• Inference engine processes the queries by
  checking rules against the customer database
• Knowledge base Rules Authorize credit only if
  customer has an active acct.
• More rules Authorize credit only if the
  customer hasnt exceeded credit limit
• More rules Authorize credit automatically if
  the customer has made five or less purchases
  today

19
Customer database

• Sam is customer acct 0000-9999
• Sam has a 5,000 credit limit
• Sam has spent 2000 in the current period
• Sam has made three transactions today

20
Expert System Response--Decision

• Yes, customer can make a 700 purchase

21
Lets look at another expert system model
22
Now, its your turn!

• In your group build an expert system. Remember,
  it is sometimes easier to work backwards from the
  solution to the problem/query