Matlin, Chapter 8 - PowerPoint PPT Presentation

1 / 26
About This Presentation
Title:

Matlin, Chapter 8

Description:

Example of importance of large knowledge base in artificial intelligence: expert ... Granny Smith. Crisp. Red. Edible. Semantic memory theories ... – PowerPoint PPT presentation

Number of Views:135
Avg rating:3.0/5.0
Slides: 27
Provided by: radf7
Category:
Tags: chapter | granny | matlin

less

Transcript and Presenter's Notes

Title: Matlin, Chapter 8


1
Matlin, Chapter 8
2
Semantic Memory
  • Semantic memory our organized knowledge base
  • Example of importance of large knowledge base in
    artificial intelligence expert systems

3
Semantic memory theories
  • Feature comparison theory
  • What is stored in memory? A feature list for
    concepts
  • Two types of features characteristic features,
    that usually occur, and defining features, that
    must occur for the concept
  • Examples what are defining and characteristic
    features of dog? Chair?

4
Semantic memory theories
  • Testing feature comparison theory sentence
    verification task
  • Example a canary is an animal
  • Dependent variable reaction time

5
Semantic memory theories
  • Prediction if two nouns have few or no features
    in common, quick RT and say no
  • If two nouns have many features in common, quick
    RT and say yes
  • If intermediate number of features in common,
    slower reaction time more variable response
    among subjects

6
Semantic memory theories
  • Typicality effect highly typical members of a
    categoryfaster reaction times on sentence
    verification task due to greater number of
    features in common with the category

7
Semantic memory theories
  • Weaknesses of feature comparison theory
  • Many categories not captured by a mere list of
    features
  • Problems determining defining features for many
    categories
  • Does not deal with relationships within categories

8
Semantic memory theories
  • Prototype theory idealized representation of a
    category probably represents average of
    exemplars in category
  • Typicalitysome exemplars more prototypical than
    others graded structure within category

9
Semantic memory theories
  • Levels of categories superordinate, basic
    level, and subordinate
  • Example motorized vehicle car Camry XLE
  • Much communication occurs at basic level useful
    levelnot too general, but not overly specific
  • Experts talk more in subordinates (specifics)

10
Semantic memory theories
  • Weaknesses of prototype theory
  • Does not deal with specific memories for
    exemplars
  • Does not describe how categories change over time
  • Not all category boundaries are as fuzzy as
    theory assumes

11
Semantic memory theories
  • Exemplar theory first item learned in a
    category becomes exemplar new items are compared
    to it (and, if similar enough, are added to
    category)
  • Additional items can be stored specifically, so
    category can have several exemplars (instead of
    just one main prototype)

12
Semantic memory theories
  • Compromise possibilities
  • Maybe small categories use exemplars and large
    categories use prototypes
  • Maybe we use both, but with different brain
    hemispheres (left for prototypes, right for
    exemplars, according to some research)

13
Semantic memory theories
  • Network theoryCollins Loftus
  • Links among nodes

14
Edible
Nutritious
Red
Fruit
Seed bearing
Crisp
Apple
Sphere-shaped
Granny Smith
Golden Delicious
15
Semantic memory theories
  • Frequency of use (how often nodes activated
    together) determines strength of link
  • Strength of link represented in network by length
    of line between nodes
  • Typicality more typical concepts used in
    connection with category more often than less
    typical concepts

16
Semantic memory theories
  • Spreading activation an activated node causes
    other nodes to which it is linked to become
    activated
  • Activation gets weaker as links are more distant
  • Can conceptualize stream of consciousness as
    spreading activation

17
Semantic memory theories
  • Andersons ACT network theory similar to one by
    Collins Loftus, but node represents a
    proposition instead of just a concept, and
    nature of relationship between nodes is specified
  • His theory actually is much more complicated, but
    beyond the scope of this book or course

18
Semantic memory theories
  • PDP (parallel distributed processing) theory
  • Neural network weighted connections similar to
    what other network theories are saying can be
    used to create simulations of knowledge networks,
    as well as useful organized knowledge bases

19
PDP example
Marti
Joe
Sam
Liberal
Engineering
Conservative
Art
Senior
Junior
Soph
Psych
20
Characteristics of PDP
  • Parallel searchinfo flows through network in
    parallel
  • Partial info still allows computation to occur
  • Spontaneous generalizations conclusion about a
    general category based on specific info in
    network (if one example of psychology major is
    liberal, might generalize that most or all psych
    majors are liberal
  • Default assignment conclusion about member of
    category due to category knowledge

21
Schemas and scripts
  • Schema any organized set of knowledge
  • Script familiar activity that usually occurs in
    a fixed sequence, involves certain actors in
    certain settings
  • Examples restaurant, going to doctor, grocery
    shopping
  • Knowing schema or script in advance aids
    comprehension

22
  • The procedure is actually quite simple. First
    you arrange
  • items into different groups. Of course one pile
    may be
  • sufficient depending on how much there is to do.
    If you have
  • to go somewhere else due to lack of facilities
    that is the next
  • step otherwise, you are pretty well set. It is
    important not
  • to overdo things. That is, it is better to do
    too few things at
  • once than too many. In the short run this may
    not seem important
  • but complications can easily arise. A mistake
    can be expensive
  • as well. At first, the whole procedure will seem
    complicated.
  • Soon, however, it will become just another facet
    of life. It is
  • difficult to foresee any end to the necessity for
    this task in
  • the immediate future, but then, one never can
    tell. After the
  • procedure is completed one arranges the materials
    into
  • different groups again. Then they can be put
    into their
  • appropriate places. Eventually they will be used
    once more
  • and the whole cycle will then have to be
    repeated. However,
  • that is part of life.

23
Schemas and scripts
  • Memory and scripts memory often is better for
    script-relevant events (could be using script to
    reconstruct memory)
  • Certain exceptions to the usual script can be
    remembered even better than script actions
    e.g., things that interrupt the usual script, are
    unexpected, emotional
  • Example the class when lightning struck the
    building just out the window

24
Schemas and scripts
  • False memory and scripts recall or recognition
    based on script can cause memory for
    script-relevant actions, props, etc. that were
    not part of a specific experience
  • Example even if article didnt mention paying
    cashier at restaurant, will later tend to
    remember having read it

25
Schemas and scripts
  • Bransford Franks study (The old car pulled the
    trailer up the steep hill.) shows how schemas
    were created (4 scenarios) that then shaped
    memorywe remember the schema (general
    representation) not the specific sentences
  • Inferencesschemas and scripts guide us in making
    inferences (I assume he paid the cashier thats
    the usual procedure)

26
Schemas and scripts
  • Inferences later can be remembered as having
    happened (source error)
  • Schema-related memory errors increase over time
    since the event
Write a Comment
User Comments (0)
About PowerShow.com