42 terms

Expert Systems

The Expert Systems unit from Higher Information Systems.
STUDY
PLAY
Describe Mycin in terms of domain, Category and defining characteristics.
Domain: Bacterial blood infections
Category: Diagnosis
Characterisitcs:Used certainty factors and backward chaining.
List the different possible conflict resolution strategies that exist.
Rule Ordering (first come first served)
Recency
Specificity
Refractoriness.
Describe the conflict resolution strategy Rule Ordering (First Come First Served).
Numerically, the first rule in the conflict set is fired.
Describe the conflict resolution strategy Recency.
The rule using the facts (or fact) most recently added to working memory is fired.
Describe the conflict resolution strategy Specificity.
The rule with the most conditions attached to it is fired.
Describe the conflict resolution strategy Refractoriness.
Once fired, a rule is permanently removed from the conflict set to prevent the system from going into a never ending loop.
What is the purpose of the RETE algorithm?
The RETE algorithm is used to maintain a conflict set.
It alleviates the burden on the inference engine, at the cost of using a great deal of memory.
Why are conflict resolution strategies required?
Because at any point in an expert system consultation more than one rule may be able to be fired.
A conflict resolution strategy makes the decision on which one will fire.
Describe Prospector in terms of domain, Category and defining characteristics.
Domain: Mineral geology
Category: Advice
Characteristics:Used fuzzy logic for uncertainty handling.
Output matched very closely that of a human expert.
Describe Oncocin in terms of domain, Category and defining characteristics.
Domain: Medical (cancer treatment)
Category: Planning
Characteristics:Automated knowledge acquisition by consulting domain experts using an interview session at a computer terminal.
Eliminated need for programmer to enter new knowledge.
Describe Dendral in terms of domain, Category and defining characteristics.
Domain: Molecular chemistry
Category: Classification
Characteristics:Used forward chaining.
Earliest expert system - developed over 20 person years.
Describe Internist in terms of domain, Category and defining characteristics.
Domain: Medical (diagnosis of internal diseases)
Category: Diagnosis
Characteristics:Combined forward and backward chaining for results.
Knowledge base had 10000 rules.
Describe Strips in terms of domain, Category and defining characteristics.
Domain: Robotic developments
Category: Planning
Characteristics:Not strictly an expert system.
Major influence in development of expert systems used for planning.
Describe R1/Xcon in terms of domain, Category and defining characteristics.
Domain: Computer configuration
Category: Planning
Characteristics:Did not require certainty handling.
mainly used specificity conflict resolution strategy.
Describe OPS5 in terms of domain, Category and defining characteristics.
Domain: Production systems
Category: Not an actual expert system
Characteristics:Introduced use of RETE algorithm.
Implemented using LISP programming language.
What are the 4 categories of expert system?
Classification
Planning
Advise
Diagnosis
What are the 3 components of an expert system?
Knowledge base
User Interface
Inference Engine
What are the 2 methods of inferencing in expert systems?
Forward chaining
Backward chaining
What is the knowledge base?
Where facts and rules are stored.
Uses production rules made up of IF statements.
Where programmer writes the code for an expert system.
What is the user interface?
Where users answer question, receive advise, and justification of advise in an expert system.
What methods of justification does an expert system provide?
How - Explanation of how the system came to provide a particular piece of advise.

Why - Explanation of why a particular question is being asked during a consultation.
In terms of inferencing, which type of expert system requires conflict resolution strategies?
A forward chaining system.
What is the inference engine?
The 'brains' of an expert system.
Applies the facts to the rules.
Decides which questions to ask via user interface.
Generates the advise at appropriate time.
Invisible to the user.
What is the purpose of the working memory?
Holds facts that are generated during a consultation.
Used by the inference engine to determine questions to ask and advise to provide.
What is an expert system shell?
A program with a ready made user interface, an inference engine and an empty knowledge base.
Saves time for programmer in building user interface.
Means all systems produce will look the same.
What are the functions of an expert system?
To provide advise an a particular domain (an area of expertise).
To handle uncertainty in both user input and advise provided.
To justify why a question is asked and how advice was provided.
What different methods are used to represent domain knowledge, before coding an expert system?
Decision trees
Factor tables
Attribute-Value pairs
What are the advantages of using certainty factors?
Allows user to realise an element of risk is associated with taking advise.
Allows user to express their degree of confidence in answers provided to questions.
Allows programmer to 'cover their back' in case of blame being apportioned is advise proves 'wrong'.
What factors might suggest a reason for developing an expert system.
Easy access to human domain experts.
Rarity of information known in a specific domain.
How is information extracted from an expert system (shell)?

(Shell - relational database comparison)
A consultation takes place.
User is asked questions and provides answers.
When sufficient information is gathered, advice is generated.
How is data stored in an expert system (shell)?

(Shell - relational database comparison)
Data is found in knowledge base in production rules.
(IF commands using AND and perhaps OR operators).
Simple facts can also be stored here.
How is information extracted from an relational database?

(Shell - relational database comparison)
A query is performed.
User enters search criteria in one or more fields.
Results based on query are returned as a series of records (perhaps viewed in a report).
How is data stored in a relational database?

(Shell - relational database comparison)
Data is stored in a series of linked tables, using fields and records.
What are the advantages of using an expert system?
Domain knowledge is preserved when a human expert dies.
Can combine knowledge of multiple domain experts.
Can provide a valuable second opinion when required.
Allows for consultation anytime night or day.
What are the design stages in creating an expert system?
Knowledge acquisition
Knowledge representation
System validation
What are the disadvantages of using an expert system?
High maintenance costs.
Advice restricted to a single domain.
Lack of common sense.
Goes out of date if not updated.
What happens during knowledge acquisition?
Domain knowledge is taken from human expert (perhaps by interview) by knowledge engineer.
What happens during knowledge representation?
Information taken from human expert represented as decision tree, factor table, attribute-value pair or similar.
Information can then be coded in an appropriare knowledge representation language.
What happens during system validation?
Expert systems is tested to ensure no faults are present.
Should involve the use of human expert, programmer and test users for best results.
What is the role of a knowledge engineer in creating an expert system?
Takes expert knowledge from expert (via interview, say) and converts to a form (such as a decision tree) that makes it easier for a program to code the system.
What is the role of an expert in creating an expert system?
Provides the expert knowledge to be added to system.
Used in testing to ensure system is error-free.
What is the role of a programmer in creating an expert system?
Codes the system in an appropriately chosen knowledge representation language (KRL).
Tests the system works as appropriate.
YOU MIGHT ALSO LIKE...