COM362 Knowledge Engineering

1
Exam Revision

• John MacIntyre
• 0191 515 3778
• john.macintyre_at_sunderland.ac.uk

2
Exam Structure

• Answer any 5 from 8 questions
• All questions worth 20 marks
• Questions divided into parts
• First part easy, bullet point answers
• Second part a little more detailed, but could
  still be answered with bullet points
• Third part requires demonstration of critical
  understanding of theory and how it can be applied

3
Points to Remember

• Use illustrations to help your answers!
• Look at how the marks are divided and allocate
  time accordingly, eg
• Three hours 180 mins
• Subtract 15 mins for reading paper, 15 mins for
  review at end 150 mins
• 30 mins per question
• Generally three parts, say 6, 6, 8 marks
• Perhaps 8, 8, 14 mins?

4
Example Q1 Part (a)

• Explain the difference between data, information,
  and knowledge. Use examples to illustrate your
  answer. (6 marks)

5
Q1a Issues to Address

• Give a definition of knowledge!
• Stress that data are the raw values information
  is the processed data knowledge is the deeper
  understanding of the information
• Use examples!
• Eg data - 5 deg C, information - it is cold,
  knowledge - because it is cold need to wear warm
  clothes.

6
Q1a Issues to Address

• Therefore knowledge includes both data and
  information
• Mention different types of knowledge
• Answers have to be brief - dont write an essay
  for 6 marks!

7
Example Q1 Part (b)

• Explain what the term Ontology means in terms
  of Knowledge Engineering. Use examples to
  illustrate your answer. (6 marks)

8
Q1b Issues to Address

• Defines vocabulary (set of standard terms),
  structure of statements, and semantic
  interpretation of terms.
• Mention different types of ontology, eg
• Terminological Ontology, Information Ontology,
  Knowledge Modelling Ontology, Domain Ontology etc
• Use examples ship design ontology, enterprise
  ontology, electrical network ontology etc

9
Example Q1 Part (c)

• You have been asked to head a team of knowledge
  engineers developing a large knowledge-based
  system to assist ship designers. You must
  minimise the amount of time taken to develop the
  system. Discuss the approach you would take to
  knowledge modelling, with specific reference to
  standardisation, compatability and software
  re-use. (8 marks)

10
Q1c Issues to Address

• Discussion of the problems of working with large
  teams of engineers and need for standardisation
  of design and modelling approach
• Development of Knowledge Level Model, and the use
  of Knowledge Modelling Ontologies for
  standardisation of design
• Discussion of different types of Ontologies, such
  as Domain and Problem Solving, and the fact that
  these are re-usable components

11
Q1c Issues to Address

• Use of existing components from an appropriate
  ontology to shorten development time and give
  compatability where a team of engineers are
  developing code.
• Additional marks for discussion of Knowledge
  Level Representation Languages, and how they are
  used to express both Knowledge Level Models and
  Knowledge Modelling Components.

12
Example Q2 Part (a)

• Describe the processes of forward and backward
  chaining. Use examples to illustrate your answer.
  (6 marks)

13
Q2a Issues to Address

• FC enter data, fire FC rules, infer new values
  from rule actions, add new values to knowledge
  base, fire all rules which can now be fired,
  repeat until all rules which can fire have been
  fired. Therefore DATA driven.
• BC state primary goal to be satisfied, find
  rules to satisfy it, fire rules and source
  sub-goals needed to satisfy those rules, get data
  at run-time if needed to resolve goals, repeat
  until all sub-goals and consequently primary goal
  are satisfied. Therefore GOAL driven.

14
Example Q2 Part (b)

• Discuss the advantages and disadvantages of a
  backward chaining approach to the problem of
  fault diagnosis for automobiles. (6 marks)

15
Q2b Issues to Address

• Advantages goal driven problem (specify fault),
  allows interaction with the user at run-time to
  get necessary information will produce a
  specific outcome.
• Disadvantages does not produce all possible
  outcomes.

16
Example Q2 Part (c)

• You are designing a knowledge-based system for
  decision support in a nuclear power station The
  plant operators wish to know everything possible
  about the state of the plant at any given time.
  Which inference mechanism would you use? Explain
  your choice. Discuss any other important issues
  for the knowledge engineer in designing safety
  critical systems. (8 marks)

17
Q2c Issues to Address

• Best choice should be forward chaining, with
  justification on the need to know everything
  possible from the given data at any time
• Some latitude if backward chaining chosen on the
  basis of the need for specific fault information

18
Q2c Issues to Address

• Safety critical issues
• auditability and traceability
• need for high accuracy and reliability
• problems with testing and validating large
  knowledge-bases
• possible conflicts
• human issues such as interface

19
General Exam Technique

• READ THE QUESTION!!
• Allocate sufficient time
• Bullet points for shorter parts
• More thoughtful discussion for longer parts
• Review answers if you have time left
• Use illustrations
• Use examples
• If you struggle for time - put SOMETHING down -
  you WILL pick up marks for this!