Knowledge Representation in AI

Introduction
• Artificial Intelligence (AI) is concerned with developing computer programs
that can perform tasks that require human intelligence. One of the essential
aspects of AI is knowledge representation, which involves capturing and
storing human knowledge in a way that machines can understand and use.
• AI aims to replicate human intelligence in machines to perform complex tasks,
including perception, reasoning, decision-making, and problem-solving.
• However, achieving this goal requires machines to have access to human
knowledge and use it to perform these tasks. Knowledge representation,
which refers to the techniques of capturing and storing human knowledge in a
way that machines can understand and use, is a critical component of AI.
• Knowledge representation is crucial in AI because it helps machines reason,
make decisions, and solve problems like humans.
Cont…
• Knowledge lays the background for two possible outcomes:
i. Gather possible sources of actions to ascertain if a particular course
of action will give the desired result.
ii. Use this judgment to determine how the action will be carried out.
• The theme of intelligence is knowledge. It is needed for the following
purposes:
i. Use, or understand, a language.
ii. Take decisions.
iii. Recognize objects.
iv. Interpret situations.
v. Plan strategies.
Cont…
• The core concepts of research in Artificial Intelligence (AI) are knowledge
representation and knowledge engineering. Objects, categories, properties
and relations between objects; situations, states, events and time; causes
and effects.
• In artificial intelligence research, commonsense knowledge is the collection
of facts and information that an ordinary person is expected to know.

• knowledge is another term for data, but knowledge is more than data or
information. Data is raw facts derived at by discovering and gathering of raw
information and researching on that information. This data is converted into
information by organizing it such that it can be made easy to draw
conclusions.
Cont…

• An AI system must be capable of :

i. Storing knowledge
ii. Applying knowledge that is preserved for solving problems and
iii. Acquiring knowledge with the help of experience.
Cont…
• five basic types of artifacts of knowledge:
(i) Facts: These are specific and unique data or instances.
(ii) Concepts: These are classes of items, words or ideas identified by a
common name, and include multiple specific examples and share
common features. Concepts are of types: concrete and abstract.
(iii) Processes: These are sequences of events or activities that
describe the working of things, rather than their functioning.
Processes are of two types: business processes, which describe the
workflow, and technical processes, which describe how things
work. In general, processes determine how anything works.
Cont…

iv. Procedures: These are a series of step-by-step actions and decisions

that cause a task to be achieved. Actions are of two types: linear and
branched.

v. Principles: These are guidelines, rules and parameters that

determine the artifacts of knowledge. They encompass both what
should be done as well as what should not be done. Principles enable
one to make predictions and deduce implications. They are the
fundamental building blocks of causal and/or theoretical models.
Two types of knowledge
• Descriptive knowledge (also known as declarative knowledge or
propositional knowledge) is the category of knowledge expressed in
declarative sentences or indicative propositions. This distinguishes
descriptive knowledge from what is commonly known as “know-
how”, or procedural knowledge (the knowledge of how, and especially
how best, to perform some task), and “knowing of”, or knowledge by
acquaintance (the knowledge of something’s existence).
• Declarative models are representations of objects and events and
their relation with other objects and events. They focus on the ‘why’
rather than the ‘how’. It allows us to think and talk about the world.
Cont…
• The special features of declarative knowledge are as follows:
• It involves the knowledge or possession of information that is
objective: either true or false.
• It is assertion oriented.
• It identifies objects and events by specifying its characteristic
properties.
• It lays stress upon the properties and not to the actions needed to
obtain a result.
• Its models include propositions and schemas.
Cont…
• Declarative knowledge is further divided into:
i. Episodic Knowledge: It refers to memory for certain episodes, such
as the context of where, when, with whom, usually measured by
accuracy measures.
ii. Semantic Knowledge: It refers to the memory for worldly
knowledge, facts, meaning of words and so on. For instance,
semantic knowledge refers to knowing that alphabetically the first
month of the year is April but chronologically it is January
Procedural Knowledge
• Procedural knowledge involves knowledge of the following aspects:
i. Formal language
ii. Symbolic representations
iii. Knowledge of rules, algorithms, and procedures.

Procedural knowledge, also known as know-how, involves how to

perform a task.

This knowledge focuses on jobs that must be performed to attain a

certain target or goal..
Advantages of procedural knowledge

• Involvement of hands-on experience

• Practice at solving problems
• Understanding the limitations of a specific solution
• Representation of heuristic knowledge
• Facilitation of default reasoning
• Enable modeling of side effects of actions
Disadvantages of procedural knowledge
• Not all cases may be represented.
• Not all deductions may be correct.
• Changes in knowledge base might have far-reaching effects.
• Controlling information is cumbersome.
• Its job dependence nature tends to be less general than declarative
knowledge. For example, computer experts proficient in more than
one language can use their knowledge about a computer algorithm in
multiple languages while a programmer proficient only Visual Basic
might know about a specific implementation of that algorithm written
only in Visual Basic.
Approaches to AI Goals
• There are four approaches to the goals of AI:
(1) Computer systems that act like humans
(2) Programs that simulate the human mind
(3) Knowledge representation and mechanistic reasoning
(4) Intelligent or rational agent design.
The first two approaches focus on studying humans and how they
solve problems,
while the latter two approaches focus on studying real-world
problems and developing rational solutions regardless of how a
human would solve the same problems.
Knowledge Progression
various kinds of knowledge that might need to
represent AI systems:
• Objects Objects are defined through facts. E.g.: Guitars with strings
and trumpets are brass instruments.
• Events Events are actions. E.g.: Vinay played the guitar at the farewell
party.
• Performance Playing the guitar involves the behaviour of the
knowledge about how to do things.
• Meta-knowledge Knowledge about what we know. To solve problems
in AI, we must represent knowledge and must deal with the entities.
• Facts Facts are truths about the real world on what we represent.
This can be considered as knowledge level
Knowledge Model
Issues in Knowledge Representation
• Important Attributes: Are there any attributes that occur in many
different types of problem? If yes, we need to ensure they are
handled appropriately in each of the mechanisms we propose.

• Relationships: What about the relationship between the attributes of

an object, such as, inverses, existence, techniques for reasoning about
values and single valued attributes?

• Granularity: It indicates the level at which knowledge should be

represented and the primitives.
Methods of Knowledge Representation

• Logic
• Semantic network
• Frames
• Production rules
• Scripts
Framework of Knowledge Representation

The steps are:

• The informal formalism of the problem takes place first.
• It is then represented formally and the computer produces an output.
• This output can then be represented in a informally described solution that the
user understands or checks for consistency.
Problem solving requires the following:
• Formal knowledge representation
• Conversion of informal (implicit) knowledge to formal (explicit) know
Knowledge and Representation

• Knowledge is a description of the world; it determines a system’s

competence by what it knows.

• Representation is the way knowledge is encoded; it defines the

system’s performance in doing something.
Cont…
In simple words, we:
• Need to know about things we want to represent, and
• need some means by which we can manipulate things.
i. Objects - Facts about objects in the domain.
ii. Events - Actions that occur in the domain.
iii. Performance - Knowledge about how to do things.
• Know things to represent
i. Meta-knowledge - Knowledge about what we know.
• Need means to manipulate.
i. Requires some formalism - To what we represent.
Representation of Facts
• Representations of facts are those which we manipulate. This can be
regarded as the symbol level since we normally define the
representation in terms of symbols that can be manipulated by
programs. We can structure these entities in two levels:
• Knowledge Level: At which the facts are going to be described. The
Symbol Level: At which representations of objects are going to be
defined in terms of symbols that can be manipulated in programs.
Properties for Knowledge Representation Systems

Knowledge representation systems possess the following properties:

• Representational adequacy
• Inferential adequacy
• Inferential efficiency
• Acquisitional efficiency
• Well-defined syntax and semantics
• Naturalness
• Frame problem
Cont…
• Representational Adequacy: A knowledge representation scheme must
be able to actually represent the knowledge appropriate to our problem.
• Inferential Adequacy: A knowledge representation scheme must allow
us to make new inference from old knowledge. It means the ability to
manipulate the knowledge represented to produce new knowledge
corresponding to that inferred from the original.
It must make inferences that are as follows:
• Sound – The new knowledge actually does follow from the old
knowledge.
• Complete – It should make all the right inferences. Here soundness is
usually easy, but completeness is very hard
Cont…

• Naturalness:
• A knowledge representation scheme should closely correspond to our
way of thinking, reading and writing.
• A knowledge representation scheme should allow a knowledge
engineer to read and check the knowledge base.
• Simply we can define knowledge representation as knowledge
representation is the problem of representing what the computer
knows.
Cont…

• Frame Problem : A frame problem is a problem of representing the

facts that change as well as those that do not change.
• For example, consider a table with a plant on it under a window.
Suppose we move it to the center of the room. Here we must infer
that plant is now in the center but the window is not.
• Frame axioms are used to describe all the things that do not change
when an operator is applied in one state to go to another state say
n + 1.
Cont…

• Inferential Efficiency:
• A knowledge representation scheme should be tractable, i.e., make
inferences in reasonable time.
• We have to use knowledge representation schemes tailored to
problem domain, i.e., less general but more efficient.
• Acquisitional Efficiency : Acquire new knowledge using automatic
methods wherever possible rather than on human intervention.
• A knowledge representation scheme should have well-defined
syntax. A knowledge representation scheme should have well defined
semantics
Well-defined syntax and semantics
• It should be possible to tell:
• Whether any construction is ‘grammatically correct’.
• How to read any particular construction, i.e., no ambiguity.
• Thus a knowledge representation scheme should have well-defined
syntax. It should be possible to precisely determine for any given
construction, exactly what its meaning is.
• Thus a knowledge representation scheme should have well defined
semantics. Here syntax is easy, but semantics is hard for a knowledge
representation scheme.
Approaches to knowledge representation
1) Simple relational knowledge:
• It is the simplest way to storing facts is to use a relational method.
• This representation gives little opportunity for inference, but it can be
used as the knowledge basis for inference engines.
• Simple way to store facts.
• Each fact about a set of objects is set out systematically in columns.
• Little opportunity for inference.
• Knowledge basis for inference engines.
• The representation of the knowledge in database form is popular.
Cont….

We can ask question like this?

1) Who is dead?
2) Who plays jazz?
● This sort of representation is used in database system.
Cont…

2) Inheritable knowledge
• The knowledge elements inherit attributes from their parents.
• The knowledge is embodied in the design hierarchies found in the
functional, physical and process domains.
• Within the hierarchy, elements inherit attributes from their parents,
but in many cases not all attributes of the parent elements be
prescribed to the child elements.
Cont…
Algorithm: Property Inheritance

1. Find the object in the knowledge base.

2. If there is a value for the attribute report it.
3. Otherwise look for a value of instance if none fail.
4. Otherwise go to that node and find a value for the attribute and
then report it.
5. Otherwise search through using isa until a value is found for the
attribute.
Techniques of Knowledge Representation in AI

• Propositional Logic, also known as propositional calculus or statement logic, is a

formal system of logic that deals with the relationships between propositions, which
are statements that are either true or false.

Propositional logic is based on the Boolean system, which means that

propositions are evaluated as either true or false. In propositional
logic, propositions are combined using logical connectives such as
"and", "or", and "not", and the resulting compound propositions can
also be evaluated as true or false based on the truth values of their |
component propositions.
Cont…

• First-order logic (FOL), also known as first-order predicate calculus

(FOPC) or first-order logic with identity, is an extension of
propositional logic that allows for the representation of more complex
relationships between objects.

• In FOL, propositions are constructed using predicates, which are

statements that describe properties or relations between objects, and
quantifiers, which specify the scope of the variables in the
proposition.
Cont…
• There are several advantages to using logical representation, such as
its ability to facilitate logical reasoning and serve as the foundation for
programming languages.

• However, there are also some limitations and challenges associated

with this method. One disadvantage is that logical representations
can be restrictive and difficult to work with. Additionally, this
approach may not be very intuitive, and the process of inference may
not always be efficient.
Cycle of Knowledge Representation in AI

• Perception
Perception is the process of acquiring information from the environment through
sensors. This information is then processed and interpreted to generate
knowledge.
• Learning
Learning is the process of acquiring new knowledge from experience. This can be
achieved through supervised learning, unsupervised learning, or reinforcement
learning.
• Knowledge Representation & Reasoning
Knowledge representation and reasoning is the stage where acquired knowledge
is transformed into a form that can be processed by machines. This involves
choosing an appropriate KR technique and representing knowledge using that
technique. Reasoning involves using the knowledge represented to
draw inferences and make decisions.
Logical Eqivalences
Logical Equivalences Involving Conditional Statement
Logical Equivalences using Biconditional Statements
Knowledge vs Intelligence

• Knowledge and intelligence are closely related concepts, but they are
not the same thing. Knowledge is information that is acquired through
experience or education. Intelligence is the ability to learn, reason,
and solve problems. Knowledge is necessary for intelligence, but it is
not sufficient.
• Intelligence involves the ability to use the techniques of knowledge
representation flexibly and adaptively. This requires not only acquiring
knowledge but also being able to reason with it, apply it to new
situations, and use it to solve problems. The ability to learn from
experience and adapt to new situations is a key aspect of intelligence.
Cont…

• For instance, techniques of knowledge representation in AI can help

AI systems learn and reason about language, enabling them to
communicate effectively with humans. They can also be used to
represent knowledge about a particular domain, such as medicine or
finance, allowing AI systems to reason about complex problems in
those domains.
• However, techniques of knowledge representation in AI alone are not
sufficient to create truly intelligent AI systems. Intelligence also
involves the ability to learn from experience and adapt to new
situations, which requires the use of machine learning algorithms and
other advanced AI techniques.
Universal and Existential Quantifiers in Predicate Logic
Exercise: To convert following sentences into 1 st order predicate logic

1. Pratap was a man.

2. Pratap was a Rajpoot.
3. All Rajpoots were Indian.
4. Rana was a ruler.
5. All Indians were either loyal to Rana or hated him.
6. Everybody is loyal to somebody.
7. People try to kill the rulers to whom they are not loyal.
8. Pratap tried to kill Rana.