Knowledge Representation

MODULE 8: KNOWLEDGE REPRESENTATION - PART 1

Concepts and Categories

  • Definitions:

    • Category: A class of similar things (objects or entities) that share essential characteristics.

    • Concept: A mental representation of an object, event, or pattern; often used interchangeably with 'category.'

  • Example:

    • All dogs belong to the category of dogs, while our mental representation of a dog is encapsulated in the concept of a dog.

Semantic Memory and Organization of Knowledge

  • The organization of knowledge about a particular object/event relates to our semantic memory, which is crucial in processing information about concepts.

ORGANIZATION OF CONCEPT KNOWLEDGE: THEORIES AND FINDINGS

Theories of Categorization

  1. Classical View:

    • Categories defined by a specific set of necessary features.

    • For example, a lawyer is defined as a human, adult, trained in law, who handles legal matters.

  2. Prototype Theory:

    • Introduced by Eleanor Rosch (1978).

    • Categories characterized by summary representations known as prototypes.

  3. Exemplar Theory:

    • Retains information about various individual instances of a category.

  4. Knowledge-Based Approach and Schemata

    • Emphasizes the role of broader knowledge in determining category membership.

Problems with Definitions

Definition Limitations

  • Definition-based categorization struggles with exceptions.

    • Example: A definition can consistently create categories, yet exceptions exist (e.g., penguins are birds but do not fit the typical image of a bird).

Typicality Effects

  • Certain members of categories appear more typical than others:

    • Study Example:

    • “Robins are birds” is verified faster than “Penguins are birds” illustrating typicality effects.

    • Individuals often name forgettable category members, starting with the more typical members instead of less typical ones.

Learning Check

Conceptual Question: Typicality effects challenge the classical view because:

  • A. Penguins should not be considered birds.

  • B. Category definitions imply equal typicality of all members.

  • C. Definitions of categories are universally applicable.

  • D. Typicality effects pose no issues for the classical view.

MODULE 8: KNOWLEDGE REPRESENTATION - PART 2

Prototype Theory

  • Prototypes defined as category characteristics with varying weights for certain features.

    • Heavily weighted features for birds include size, ability to fly, beak.

    • Less critical features include color.

Nature of Prototypes

  • Unlike the classical view, not all features are necessary for classification.

  • Categories may have fuzzy boundaries and graded membership, which can explain typicality effects.

Levels of Categorization

  • Basic Level: Optimal level of categorization that's neither too general nor specific,

    • Example: "Chair" is the basic level of the superordinate category "furniture."

Category Levels Overview

  • Example Structure:

    • Superordinate: Musical Instrument

    • Basic Level: Guitar

    • Subordinate: Classical guitar

Learning Check

  • Identify the correct order of category lists:

    • A. Dog, Animal, Poodle

    • B. Petunia, Flower, Plant

    • C. Furniture, Lamp, Table-Lamp

    • D. Gala apple, Apple, Fruit

MODULE 8: KNOWLEDGE REPRESENTATION - PART 3

Exemplar/Instance Theory

  • Storing individual instances rather than relying solely on prototypes offers a more nuanced understanding of categories.

  • Category membership determined by similarity to stored exemplars.

Advantages and Disadvantages

  • Pros: Retains variability information, allows richer concepts.

  • Cons: Hierarchical structures may be problematic.

Typicality Effects in Theories

  • Both prototype and exemplar theories can explain typicality effects:

  • Activation of the prototype along with frequent exemplars leads to quicker categorization.

Learning Check

  • A significant difference between prototype and exemplar theories is:

  • A. Prototype theory is solely responsible for typicality effects.

  • B. Exemplar theory states that specific instances are encoded distinctly.

Challenges with Categorization via Similarity

  • Situations exist where categorization does not rely solely on resemblance, evidenced through novel examples.

Knowledge-Based Approach

  • Broader knowledge aids in categorization, contributing to a comprehensive understanding of each category's limitations.

Schemata and Scripts

  • Integrates the prototype and knowledge-based approaches with organized information structures, including scripts that represent routine events.

    • Example: Schema for "dog" incorporates characteristics and variations, reflecting the broad understanding of the category.