Representation of knowledge
Knowledge representation promote exploration of a wide range of
phenomena and offer the strength of converging
operations—the use of multiple approaches and techniques to address a
problem. The way in which knowledge is represented profoundly influences how
effectively knowledge can be manipulated for performing any number of cognitive
tasks.
The fundamental unit of symbolic knowledge is the concept—an
idea about something that provides a means of understanding the world. One way
to organize them is by means of categories. A category is a
group of items into which different objects or concepts can be placed that
belong together because they share some common features, or because they are
all similar to a certain prototype.
·
Categories
·
Prototypes
·
Schemas
·
Scripts
·
Acquisition
of declarative & procedural knowledge
Earlier models were based on behaviorist theory. “ Stimulus-response association theory was proposed by
Clark Hull (1920). He argued that we learn to associate a particular response
(the concept) with a variety of stimuli that define the concept.”
Jerome Bruner formulated a concept formation
theory that involved cognitive processes, i.e. hypothesis testing about a
concept by making guesses about which attributes are essential for defining the
concept. Concept attainment according to Bruner et al. 1967:233)
is "the search for and listing of attributes that can be used to
distinguish exemplars from nonexemplars of various categories"
Eleanor Rosch (1978) suggested that the natural
concepts in everyday life are learned through examples rather than abstract
rules.
Anderson's Adaptive Control of Thought (ACT)
theory suggests that long-term memory is an interconnect network of
propositions (facts of concepts). Only a subset of interconnected propositions
can be activated and more connected propositions are easier to retrieve. A
concept that has many connections is elaborated.
“ Tennyson & Cocchiarella (1986) suggest a
model for concept teaching that has three stages: (1) establishing a connection
in memory between the concept to be learned and existing knowledge, (2)
improving the formation of concepts in terms of relations, and (3) facilitating
the development of classification rules. This model acknowledges the
declarative and procedural aspects of cognition.”
“ Klausmeier (1974) suggests four levels of
concept learning: (1) concrete - recall of critical attributes, (2) identity -
recall of examples, (3) classification - generalizing to new examples, and (4)
formalization - discriminating new instances.”
Using Our Minds
–
Knowing that…Declarative knowledge
–
Knowing how…Procedural knowledge
Declarative Knowledge
•
Stored in Concepts
–
A mental representation of an item and associated
knowledge and beliefs (cat, tools, furniture)
When Do We Use Concepts?
•
Create categories
•
Make inferences
•
Combine to form complex thoughts
•
For communication
Organizing Structures of Declarative Knowledge
•
Concept --Unit
of symbolic knowledge
•
Category --Rule
used to organize concepts
•
Schemas -- Framework used to organize concepts
Different Types of Concepts
•
Natural Concept
–
Occur naturally (e.g. plants, trees, cats)
•
Artifact Concept
–
Created by humans (e.g., hammers, computers)
•
Ad Hoc Concepts
–
Created individually to suit a need (things you need to
be happy, things you do to please parents)
Different Theories on Concept Organization
•
Defining Features (Classical View)
•
Prototypes
•
Exemplars
•
Hierarchically semantic networks
Categorization and
Concepts
o Basic
cognitive function is to categorize
n Use
experience to aid in future behavior and decision-making
o Cognitive
economy
o Concepts
n Mental
representation of a category serving multiple functions
Functions of Concepts
o Classification
n Determine
category membership
o Understanding,
making predictions, inference
n Once
classified one can then understand its relevant parts, know how to interact
with it, infer other properties
o Explanation
and Reasoning
n For
example, of others’ behavior
o Learning
n New
entities compared to and understood in terms of old and provide feedback for
modification
o Communication
n Shared
concepts and categorization allow for easier expression of ideas to others
Categories
o Categories
n Collection
of objects, attributes, or actions, etc.
o List
of concepts
o Hierarchy
n Set
of entities or examples picked out by the concept
o How
is experience distilled? How are functional relations established?
n Category
learning
o How
is knowledge represented in a category?
n Structure
n Schema
o General
knowledge structure that integrates objects, attributes, and actions into a
cohesive representation
n Script
n Sequence
o How
do we use categorical knowledge?
Classification
o Determining
the category membership of various things (objects, properties, abstractions
etc.)
o Allows
for treating otherwise discriminable entities as similar
n Similarity
as the organizing principle for categories and categorization
Types of Categories
n Abstract
vs. Concrete
o Love
vs. Mammal
n Hierarchical
vs. Non
o Mammal
vs. woman
n Feature-Based Categories: A Defining View
All those features are then necessary (and sufficient) to
define the category. This means that each feature is an essential element of
the category. Together, the features uniquely define the category; they are defining
features (or necessary attributes): For a thing to be an X, it must have that
feature. Otherwise, it is not an X.
Ex. Bachelor (male, unmarried, adult) The features are
each single necessary. If one feature is absent, the object cannot belong to
the category. The three features are jointly sufficient. If a person has
all three features, then he is automatically a bachelor.
Prototype Theory:
A Characteristic View Prototype Theory – grouping of things together not by
their defining features but rather by their similarity to an averaged model of
the category. Prototype is an abstract average of all objects in the category
we have encountered before. Crucial are characteristic features, which
describe (characterize or typify) the prototype but are not necessary for it.
They are commonly present in typical examples of concepts, but they are not
always present.
Ex. Prototype of a game Prototype of a bird (robin or ostrich)
Whereas a defining feature is shared by every single object in a category, a
characteristic feature need not be.
Classical concepts are categories that can be readily
defined through defining features, such as bachelor. Fuzzy concepts are categories
that cannot be so easily defined, such as game or death.
Exemplars are typical representatives of a category. Ex.
Birds, we might think not only of the prototypical songbird, which is small,
flies, builds nest, sings, and so on. We also might think of exemplars for
birds of prey, for large flightless birds, for medium-sized waterfowl, and so
on.
Schemas
•
Schemas
are models of the external world based on past experience
•
Schemas
for concepts underlying situations, events, or sequences of actions
•
Abstraction
that allows particular objects or events to be assigned to general categories
•
Organize our
knowledge
•
May include
other schemas
•
Help in
encoding, storage, and recall
•
Allows us to
make inferences
Schemas –
mental framework for organizing knowledge. It creates a meaningful structure of
related concepts.
A cognitive structure that organizes related concepts and
integrates past events. Ex. Kitchen (tells us the kind of things we might
find in a kitchen and where we might find them)
Schemas have several characteristics that ensure wide
flexibility in their use. Schemas can include other schemas. Ex. A schema for
animals includes a schema for cows, a schema for apes, and so on. Schemas
encompass typical, general facts that can vary slightly from one specific
instance to another. Schemas can vary in their degree of abstraction.
Scripts
•
Type of schema about events
•
Structure captures general
information about routine events
–
Eating in a restaurant, attending a
movie, a visiting a doctor’s office
•
Scripts have typical roles
–
(Customers, waiter, cook), (ticket
vendor, patrons, refreshments), (doctor, nurse, patient)
•
When we hear or read about a
scripted event, our knowledge of the entire script is activated
•
We can fill in or infer the scenes
and actions that are not explicitly mentioned
Script contains
information about the particular order in which things occur. Ex. Restaurant
script (coffee shop) Props: tables, a menu, food, a check, and money Roles
to be played: a customer, a waiter, a cook, a cashier, and an owner. Opening
conditions for the script: the customer is hungry, and he or she has money
Scenes: entering, ordering, eating, and exiting A set of results: the
customer has less money; the owner has more money; the customer is no longer
hungry; and sometimes the customer and the owner are pleased.
Jargon – specialized vocabulary commonly used within a
group, such as a profession or a trade. Imaging studies reveal that the
frontal and parietal lobes are involved in the generation of scripts. The
generation of scripts requires a great deal of working memory. Further script
generation involves the use of both temporal and spatial information.
Scripts enable us to use a mental framework for acting in
certain situations when we must fill in apparent gaps within a given context.
Acquisition of declarative & procedural
knowledge
Representing Procedural Knowledge
•
Serial Processing
–
Linear sequence of operations
–
Create using production rules
•
If – then rules
–
If sliding on ice then pump the brakes
•
Tasks may take multiple rules
–
Organized into routines and subroutines
ACT-R Model
•
Combines declarative and procedural knowledge in a
model
•
Declarative knowledge is represented in structures
called chunks defined by its type and slots
–
Type represents concepts or categories (e.g., dogs) and
slots as category attributes (e.g., color or size)
•
Anderson (1980)
–
Cognitive Stage
•
Consciously think about steps to complete task
–
Associative Stage
•
Practice the procedure
–
Autonomous Stage
•
Skill has become automatic
Squire’s
Non-declarative Knowledge
•
Procedural knowledge
•
Associative conditioning
–
Classical and operant conditioning
•
Simple nonassociative knowledge
–
Habituation
–
Sensitization
•
Priming
Support for Squire’s Taxonomy
•
Basil Ganglia damage
•
Examine Parkinson’s and early Huntington disease
patients
•
No apparent amnesia (declarative memory intact)
•
Problems with procedural memory
–
Perceptual motor learning
–
Habits & skills
•
Just one example of variety of studies with humans and
animals have supported Squire’s taxonomy
Two Types of Priming
•
Semantic priming
–
Meaning is primed
–
Remember Nurse-Doctor study?
•
Repetition priming
–
Prior exposure primes same items seen later
Connectionist Model
•
Parallel processing
–
Multiple operations occur simultaneously
•
Parallel Distributed Processing (PDP) models
–
Goal is to model information as it is represented in
the brain
The PDP Model
•
The representation of information is distributed
•
Knowledge for specific things are not stored
explicitly, but stored in the activations of patterns among units
•
Learning occurs with changes in connection strength by
experience
•
Units send excitatory and inhibitory signals to other
unit
No comments:
Post a Comment