Knowledge Representation for Natural Language Competence - PowerPoint PPT Presentation

About This Presentation
Title:

Knowledge Representation for Natural Language Competence

Description:

Did Lois Lane see Superman? I don't know. : Did Lois Lane see Clark Kent? ... Superman went to the morning star. I understand that Superman went to Venus. ... – PowerPoint PPT presentation

Number of Views:16
Avg rating:3.0/5.0

less

Transcript and Presenter's Notes

Title: Knowledge Representation for Natural Language Competence


1
Knowledge Representation for Natural Language
Competence
  • Stuart C. Shapiro
  • Department of Computer Science and Engineering
  • and Center for Cognitive Science
  • State University of New York at Buffalo
  • shapiro_at_cse.buffalo.edu

2
Motivation
  • Deep understanding of NL texts requires a
    Knowledge Representation Reasoning
    formalism/system.
  • A variety of logic.
  • But not the logic for metamathematics.

3
Preview
  • I will discuss several issues in KRR for NL
    Competence,
  • Illustrated by interactions with a computational
    agent.
  • Agent Cassie.
  • KRR system SNePS.
  • Interactions are transcribed from consecutive
    interactions in a single session.

4
Interaction with Cassie
(Current) Set of Beliefs SNePS
English (Statement, Question, Command)
Reasoning Clarification Dialogue Looking in World
GATN Parser
(Updated) Set of Beliefs SNePS
Actions SNeRE
(New Belief) SNePS
Answer SNIP
GATN Generator
Reasoning
English sentence expressing new belief
answering question reporting actions
5
Outline
  • Introduction
  • Intensional Entities
  • Complex Categories
  • Possession I
  • Propositions about Propositions
  • Possession II
  • Summary

6
Intensional Entities 1
  • Rather than represent objects in the world,
    represent mental entities.
  • Includes Imaginary and Fictional Entities.
  • Multiple mental entities may correspond to one
    world object.
  • Intensional entities may be co-extensional.
  • But must be kept separate.

7
Intensional Entities 2
  • The morning star is the evening star.
  • I understand that the morning star is the
    evening star.
  • The evening star is Venus.
  • I understand that Venus is the evening star.
  • Clark Kent is Superman.
  • I understand that Superman is Clark Kent.

8
Intensional Entities 3
  • Lois Lane saw Clark Kent.
  • I understand that Lois Lane saw Clark Kent.
  • Did Lois Lane see Superman?
  • I don't know.
  • Did Lois Lane see Clark Kent?
  • Yes, Lois Lane saw Clark Kent.

Note Open World Assumption.
9
Intensional Entities 4
  • Superman went to the morning star.
  • I understand that Superman went to Venus.
  • Did Clark Kent go to Venus?
  • Yes, Superman went to Venus.

10
Intensional Entities 5
  • Buck Rogers went to the evening star.
  • I understand that Buck Rogers went to Venus.
  • Who went to Venus?
  • Buck Rogers went to Venus
  • and Superman went to Venus.

11
Intensional Entities 6
The evening star
The morning star
Venus
Go to
Go to
Clark Kent
Superman
Buck Rogers
See
Lois Lane
12
Outline
  • Introduction
  • Intensional Entities
  • Complex Categories
  • Possession I
  • Propositions about Propositions
  • Possession II
  • Summary

13
Complex Categories 1
  • Noun Phrases
  • ltDetgt N Adj N
  • Understanding of the modification must be left to
    reasoning.
  • Example
  • orange juice seat
  • Representation must be left vague.

14
Complex Categories 2
  • Kevin went to the orange juice seat.
  • I understand that Kevin went to the orange juice
    seat.
  • Did Kevin go to a seat?
  • Yes, Kevin went to the orange juice seat.

15
Complex Categories 3
  • Pat is an excellent teacher.
  • I understand that Pat is an excellent teacher.
  • Is Pat a teacher?
  • Yes, Pat is a teacher.
  • Lucy is a former teacher.
  • I understand that Lucy is a former teacher.

16
Complex Categories 4
  • former' is a negative adjective.
  • I understand that former' is a negative
    adjective.
  • Is Lucy a teacher?
  • No, Lucy is not a teacher.

17
PseudoRepresentation of Complex Categories
  • Isa(B30, CompCat(orange, CompCat(juice, seat)))
  • Isa(Pat, CompCat(excellent, teacher))
  • Isa(Lucy, CompCat(former, teacher))

18
Outline
  • Introduction
  • Intensional Entities
  • Complex Categories
  • Possession I
  • Propositions about Propositions
  • Possession II
  • Summary

19
Possession I.1
  • One mans meat is another mans poison.

20
Possession I.2
  • Richard's meat is Henry's poison.
  • I understand that Henry's poison is Richard's
    meat.
  • Edward ate Richard's meat.
  • I understand that Edward ate Richard's meat.
  • Did Edward eat Henry's poison?
  • Yes, Edward ate Henry's poison.

21
Possession I.3
  • Did Edward eat Henrys meat?
  • I dont know.
  • Did Edward eat Richard's poison?
  • I dont know.

Moral Possession is a three-place relation.
22
PseudoRepresentation of Possession
  • Has(Richard, meat, B35)
  • Has(Henry, poison, B37)
  • Equiv(B35, B37)

23
Outline
  • Introduction
  • Intensional Entities
  • Complex Categories
  • Possession I
  • Propositions about Propositions
  • Possession II
  • Summary

24
Propositions about Propositions 1
  • Propositions are first-class mental entities.
  • They can be discussed, just like other mental
    entities.
  • And must be represented like other mental
    entities.

25
Propositions about Propositions 2
  • That Bill is sweet is Mary's favorite
    proposition.
  • I understand that Mary's favorite proposition is
    that Bill is sweet.
  • Mike believes Mary's favorite proposition.
  • I understand that Mike believes that Bill is
    sweet.

26
Propositions about Propositions 3
  • That Mary's favorite proposition is that
    Bill is sweet is cute.
  • I understand that that Mary's favorite
    proposition is that Bill is sweet is cute.

27
Representing Propositions
  • Representation of Proposition
  • Not by a Logical Sentence
  • But by a Functional Term
  • Denoting a Proposition.

28
PseudoRepresentation of Propositions about
Propositions
  • Has(Mary, CompCat(favorite, proposition),
    HasProp(Bill, sweet))
  • Believes(Mike, HasProp(Bill, sweet))
  • HasProp(Has(Mary,
  • CompCat(favorite,
    proposition), HasProp(Bill, sweet)),
  • cute)

29
Outline
  • Introduction
  • Intensional Entities
  • Complex Categories
  • Possession I
  • Propositions about Propositions
  • Possession II
  • Summary

30
Possession II.1
  • Examples from J. Lyons, Semantics I, 1977, p.
    312,
  • of inalienable possessive constructions
  • Johns right arm
  • of alienable possessive constructions
  • Johns book
  • Use vague representation
  • with later reasoning.

31
Possession II.2
  • Caren held Stu's hand.
  • I understand that Caren held Stu's hand.
  • Mary held Bill's book.
  • I understand that Mary held Bill's book.

32
Possession II.3
  • What is an inalienable possession?
  • I don't know.
  • What is an alienable possession?
  • I don't know.

33
Possession II.4
  • Hands are body parts.
  • I understand that hands are body parts.
  • Books are ownable objects.
  • I understand that books are ownable objects.

34
Possession II.5
  • What is Stu's hand?
  • Stus hand is the hand.
  • What is Bill's book?
  • Bills book is the book.

35
Possession II.6
  • What is an inalienable possession?
  • That Stu has a hand is the inalienable
    possession.
  • What is an alienable possession?
  • That Bill has a book is the alienable possession.

36
PseudoRepresentation of Possession
  • Has(Stu, hand, B47)
  • Has(Bill, book, B49)
  • Held(Caren, B47)
  • Held(Mary, B49)
  • Isa(Has(Stu, hand, B47), CompCat(inalienable,
    possession))
  • Isa(Has(Bill, book, B49), CompCat(alienable,
    possession))

37
Outline
  • Introduction
  • Intensional Entities
  • Complex Categories
  • Possession I
  • Propositions about Propositions
  • Possession II
  • Summary

38
Summary
  • Represent intensional (mental) entities.
  • Open World Assumption
  • Vague representation of complex categories.
  • Clarified by reasoning.
  • Ability to discuss words.
  • NL is its own metalanguage.
  • Possession as a three-place relation.
  • Propositions as first-class entities.
  • Vague representation of possession.
  • Clarified by reasoning.
  • Supplying taxonomy via NL inputs.

39
SNePS Research GroupCurrent Members
  • Faculty
  • Stuart C. Shapiro, Director
  • William J. Rapaport, Associate Director
  • Carl Alphonce
  • Jean-Pierre A. Koenig
  • David R. Pierce
  • Graduate Students
  • Marc Broklawski Bharat Bhushan
  • Debra T. Burhans Haythem O. Ismail
  • Frances L. Johnson John F. Santore

40
For More Information
  • URL http//www.cse.buffalo.edu/shapiro/
  • Group http//www.cse.buffalo.edu/sneps/
Write a Comment
User Comments (0)
About PowerShow.com