CSCI3406 Fuzzy Logic and Knowledge Based Systems

1
CSCI3406 Fuzzy Logic and Knowledge Based Systems
Knowledge Acquisition - II Knowledge
Representation I KBS Tutorial (patient to
medicine)
2
Introduction

• Briefly discuss general KA techniques
• Cover the different techniques used for Knowledge
  Representation
• What makes us to choose one technique or another?

3

• KA Techniques
• There have been many techniques developed to
  help elicit knowledge from expert(s). These are
  referred to as knowledge elicitation or knowledge
  acquisition (KA) techniques. For example
• Interviews (General, Focused and Structured
  Interview)
• Observation
• Protocol Analysis
• Walkthroughs
• Repertory Grids
• Computer aided Knowledge Acquisition
• Automated Rule Induction

4
Interviews (Focused and Structured
Interview) Follow the followings to conduct a
productive interview -Prepare yourself before
the interview takes place -Keep the expert on
track -Explain yourself make yourself
clear -Document all contents of the interview
(know how to take note?) Focused Interview For
this type of interview the knowledge engineer
decides before the interview exactly the
questions to be asked. It may even be appropriate
to provide questions to the expert beforehand.
When would this be most appropriate? It is
usually most useful in early interviews for
gaining a better understanding of the
domain. Structured Interview In this case the
knowledge engineer will use subject headings to
control the order in the interview. These subject
headings should be agreed with the expert at the
beginning of the interview. The expert should ask
the expert to give an overview of the topic but
the knowledge engineer should expect to interrupt
the expert where appropriate for clarification.
5
Observation It can be very useful for the
knowledge engineer to observe the expert in
action as it can often be very difficult for an
expert to explain what he/she does and indeed the
interpretation of what he/she actually does may
not be totally accurate. Protocol Analysis
Protocol analysis, typically, is the process by
where the expert is asked to "think aloud"
throughout the solving of the problem.
Walkthroughs A walkthrough is where the
knowledge engineer asks the expert to "walk"
through the job they do. The knowledge engineer
has the capability to interrupt the expert asking
why/how questions whereas in protocol analysis
interruptions should be avoided at all
costs. Repertory Grids Repertory grids are used
to identify attributes and to describe
objects. Computer Aided Knowledge Acquisition
Computers should be used effectively in the
process of acquisition of knowledge in order to
increase productivity and (sometimes) eliminate
the need for an expert.
6

• Automated Rule Induction
• This technique takes a set of examples to try and
  generate general rules
•   Advantages
• - If the knowledge is very complex and/or the
  domain is large then it may not be possible to
  develop expert systems.
•   - The developer of a system does not have to
  have all the skills of a knowledge engineer.
• - This approach may well allow for new
  knowledge to be acquired.
• Once a base set of rules have been developed
  using rule induction these can be modified by the
  knowledge engineer and expert together
• Disadvantages/Problems
•         How are the attributes chosen?
•         Algorithms are inefficient.
•         Only suitable for rule based systems
•         How do we choose the training set?
•         The algorithms cannot usually deal with
  exceptions

7
Knowledge Representation

• What is knowledge representation?
• Logic representation
• Semantic networks
• Scripts and frames

8
What is knowledge representation?

• Once we have acquired knowledge we need a way to
  represent this knowledge in a computable
  way/model.
• Definitions of Knowledge Representation (KR)
• The notation or formalism used for coding the
  knowledge to be stored in a knowledge-based
  system. www.pcai.com/web/glossary/pcai_d_f_glossar
  y.html
• The process and the result of formalization of
  knowledge in such a way that it can be used
  logically and automatically for problem solving.
  www.centc251.org/Ginfo/Glossary/tcglosk.htm
• The KR term used in artificial intelligence to
  cover the study of formalisms which model human
  forms of knowledge. www.informatics.susx.ac.uk/boo
  ks/computers-and-thought/gloss/node1.html
• KR is a formalism for representing facts and
  rules in a computer about a subject or a
  specialty. www.wiley.co.uk/college/turban/glossary
  .html
• There are two main streams in knowledge
  representation Symbolic Numeric

9
What is knowledge representation?

• It is argued that the notion can best be
  understood in terms of five distinct roles it
  plays, each crucial to the task at hand
• A knowledge representation (KR) is most
  fundamentally a surrogate, a substitute for the
  thing itself, used to enable an entity to
  determine consequences by thinking rather than
  acting, i.e., by reasoning about the world rather
  than taking action in it.
• It is a set of ontological commitments, i.e., an
  answer to the question In what terms should I
  think about the world?
• It is a fragmentary theory of intelligent
  reasoning, expressed in terms of three
  components (i) the representation's fundamental
  conception of intelligent reasoning (ii) the set
  of inferences the representation sanctions and
  (iii) the set of inferences it recommends.
• It is a medium for pragmatically efficient
  computation, i.e., the computational environment
  in which thinking is accomplished. One
  contribution to this pragmatic efficiency is
  supplied by the guidance a representation
  provides for organizing information so as to
  facilitate making the recommended inferences.
• It is a medium of human expression, i.e., a
  language in which we say things about the world

REF (must-read) Detail discussion on What is
knowledge representation? can be found on R.
Davis et al., What is a Knowledge
Representation?, AI Magazine, vol.14(1),
pp17-33, 1993 http//groups.csail.mit.edu/medg/f
tp/psz/k-rep.html
10
Logic Representation

• Logic or mathematical logic is one of the oldest
  computer modelling techniques and current
  computers are based on some logical models.
• The typical logical process consists of inputs
  that represent facts or premises and outputs that
  represent inferences or conclusions.
• Logic representation comes from mathematical
  proofs, thus, sometimes we refer to logic-based
  systems as problem solvers or automated proof
  systems.

11
Logic Representation

• There are several logical models, however, many
  of them are difficult to compute with our current
  processing resources.
• As a result, there are two main logical theories
  that are often used propositional logic and
  predicate logic (calculus).
• They are based around the idea of propositions
  and their truth value.

12
Logic Representation

• Main difference is that predicate calculus uses
  predicates, what does that mean?
• Some of the operators and notation used are

13
Prolog for Logic Representation

• A programming language (Prolog) has been
  developed to enable the implementation of logic
  models.
• Advantages of Prolog
• Very powerful and flexible,
• No need to write an inference mechanism.
•  Disadvantages of Prolog
• Difficult! And It can take a while to become
  proficient in its use.
• Very difficult to develop a good user interface.

14
Semantic Networks

• Semantic nets are basically graphic depictions of
  knowledge that show hierarchical relationships
  between objects and their attributes, concepts,
  events or actions. 
• Semantic networks are made of a number of
  circles, which are referred to as nodes, and
  links, which are referred to as arcs or
  relationships. Each node can represent an object,
  attribute, concept, event or action. These nodes
  are also interconnected by links, or arcs. These
  arcs show the relationships between the various
  objects and descriptive factors, i.e. attributes.
  Some of the most common arcs are of the is-a
  (class relation) or has-a (attribute relation)
  type.
• One of the most interesting and useful facts
  about a semantic network is that it can show
  inheritance supported by is-a links type

15
Scripts and Frames-based systems

• A frame is a relatively large block or chunk of
  knowledge about a particular object, event,
  location, situation, or other element. The frame
  describes that object in great detail (each frame
  for one object)
• A script as a knowledge representation scheme is
  similar to a frame, but instead of describing an
  object, the script describes a sequence of
  events. To describe a sequence of events, the
  script uses a series of slots containing
  information about the people, objects,
  and actions that are involved in the events. Some
  of the elements of a typical script are
• entry conditions describe situations that must
  be satisfied before events in this script can
  occur or be valid,
• props refer to objects that are used in the
  sequence of events that occur
• roles refer to the people involved in the
  script. The result is conditions that exist after
  the events in the script have occurred
• tracks refers to variations that might occur in
  a particular script
• scenes describe the actual sequence of events
  that occur.

16
Tutorial

• A group work to develop a KBS modelling a
  process from a person who feels sick to medicine
  that he needs

17
Semantic Networks
See the last weeks lecture notes and handouts
for example and further details
18
Script
See the last weeks lecture notes and handouts
for example and further details
19
Tutorial You are asked to develop a KBS system
based on Semantic Networks and/or Script type of
Knowledge representation. For this KBS, you will
be modelling a system for a patient who feels
sick and needs medicine. You will be modelling
the process from patient to medicine.
Medicine
20

• Tutorial
• For the semantic networks, you should be
  identifying nodes and links as many of them as
  possible
• For the script type of knowledge representation,
  you should be identifying entry conditions, props
  , roles, tracks, and scenes

Medicine