Harsh RB

Knowledge Representation and Reasoning Class Notes

3 min read

Unit 1: Knowledge Representation

Knowledge Representation (KRR) is the study of how beliefs, intentions, and knowledge are stored and utilized to solve complex real-life problems, such as communicating with human beings in natural language.

  • It allows machines to learn from that knowledge and behave intelligently like a human being.

Different Kinds of Knowledge in AI

  • Object
  • Events
  • Performance
  • Meta Knowledge
  • Knowledge Base

Types of Knowledge

  • Declarative Knowledge
  • Structural Knowledge
  • Procedural Knowledge

Cycles of Knowledge Representation in AI

  • Perception
  • Learning
  • KRR
  • Planning
  • Execution

Techniques of Knowledge Representation

Logical Representation

  • It decides how we construct legal sentences in logic.
  • It determines which symbols we can use in knowledge representation.

Example: With P,Q,R: Truth table

Semantic Representation

In a Semantic Network, knowledge is represented in the form of a graphical network.

  • This network consists of nodes representing objects and arcs that describe the relationships between these objects.

This representation consists of two types of relations:

  • Is-a
  • Kind-of

Example: Family Tree

Frame Representation

  • It is a record-like structure that consists of a collection of attributes and values to describe an entity in the world.

Example: Bus

Production Rule

It consists of three main parts:

  1. Set of Production Rules
  2. Working Memory
  3. Recognized Cycle

Role of Logic in Knowledge Representation and Reasoning

  • The relationship between relational language, truth conditions, and rules of inference.
  • The first knowledge representation language is a very popular logical language. It is sometimes called First-Order Logic.

It has two levels:
4. Knowledge Level – We ask questions concerning the representing language and its semantics.
5. Symbol Level – We ask questions concerning the computational aspects.

Syntax

  • Specifies a group of words or symbols that should be properly formed.
  • Example:
    • “The cat my mother loves” → Well-formed phrase
    • “The love mother cat” → Not meaningful in knowledge representation

Semantics

  • Expressions must be meaningful and represent real-world concepts.
  • Example:
    • “The hard-nosed decimal holiday” might not mean anything.
    • We need to be clear about what idea about the world is being expressed.

Pragmatics

  • Specifies how meaningful expressions in the language are to be used.
  • Example:
    • “There is someone behind you”
    • Could be a warning to be careful in some contexts.

Syntax

Logical Symbols

  • Punctuation: "", ''
  • Connectives: AND, OR, NOT
  • Quantifiers: ∃ (There Exists), ∀ (For All)
  • Variables: Changes with functions

Non-Logical Symbols

  • Functional
  • Practical Symbols

Sample Project

Home Automation System using KRR

Develop a Home Automation System that uses Knowledge Representation and Reasoning (KRR) to:

  • Understand commands and context
  • Process requests intelligently

Example Use Case:

  • Turn on Light when sunset and temperature drops below 65°F

System Architecture

  • IoT Integration
  • KRR Implementation
  • Automation Rules
  • Natural Language Processing (NLP)

Graph View

Backlinks