[PPT] - Communicat ion J uly 14, 2005 CS 486/ 686 Universit y of Wat PowerPoint Presentation

SLIDE 1

Communicat ion

J uly 14, 2005 CS 486/ 686 Universit y of Wat erloo

SLIDE 2

2

Out line

Communicat ion
Symbolic Nat ural Language Processing
Reading: R&N Sect . 22.1-22.6

SLIDE 3

3

Communicat ion

Communicat ion: int ent ional exchange of

inf ormat ion brought about by t he product ion and percept ion of signs drawn f rom shared syst em of convent ion.

Language:

– Enables us t o communicat e – I nt imat ely t ied t o t hinking

SLIDE 4

4

Turing Test

Can a comput er f ool a human t o t hink

t hat it is communicat ing wit h anot her human?

SLIDE 5

5

Speech

Speech: communicat ion act

– Talking – Wr it ing – Facial expression – Gest ure ut t erances Speaker Hearer ut t erances situation

SLIDE 6

6

Component s of Communicat ion

I nt ent ion

– Speaker S decides t hat t here is some proposit ion P wort h saying t o hearer H.

Generat ion

– Speaker plans how t o t urn proposit ion P int o an ut t erance (i.e. a sequence of words W)

Synt hesis

– Speaker produces t he physical realizat ion W’ of t he words W (i.e., vibrat ion in air, ink on paper)

SLIDE 7

7

Component s of Communicat ion

Percept ion

– Hearer perceives physical realizat ion W’ as W2 and decodes it as t he words W2 (i.e., speech recognit ion, opt ical charact er recognit ion)

Analysis

– Hearer inf ers W 2 has possible meanings P

1, P 2, …

, P

n

– Three part s:

Synt act ic int erpret at ion
Semant ic int erpret at ion
Pragmat ic int erpret at ion

SLIDE 8

8

Component s of Communicat ion

Disambiguat ion

– Hearer inf ers t hat speaker int ended t o convey P

i (where ideally P i = P).

I ncorporat ion

– Hearer decides t o believe P

i (or not ).

SLIDE 9

9

Component s of Communicat ion

"The wumpus is dead"

Article Noun The wumpus is dead Verb Adjective NP VP S Tired(Wumpus,Now) Perception: Analysis: (Parsing):

(Semantic Interpretation):

HEARER

3

L

Alive(Wumpus,S ) Incorporation: TELL( KB,

(Pragmatic Interpretation):

3

Tired(Wumpus,S )

3

L L

Disambiguation:

3

L

Alive(Wumpus,S ) Alive(Wumpus,Now) Alive(Wumpus,S )

"The wumpus is dead"

Intention: Generation: Synthesis: SPEAKER

[thaxwahmpaxsihzdehd]

Know(H, Alive(Wumpus,S ))

L

3

SLIDE 10

10

Dif f icult ies

How could communicat ion go wrong?

– I nsincer it y – Speech recognit ion errors – Ambiguous ut t er ance – Dif f erent cont ext s

SLIDE 11

11

Language

Formal language

– Set of st rings of t erminal symbols (words) – St rict rules – E.g., f irst order logic, J ava

Nat ural language

– No st rict def init ion – Chinese, Danish, English, et c.

SLIDE 12

12

Grammar

Grammar specif ies t he composit ional

st ruct ure of complex messages

Each st ring in a language can be

analyzed/ generat ed by t he grammar

A grammar is a set of rewrit e rules

– S NP VP – Ar t icle t he | a | an | …

SLIDE 13

13

Grammar Types

Regular grammar:

– nont erminal t erminal [nont erminal] – S a S – S b

Cont ext f ree grammar (CFG):

– nont erminal anyt hing – S aSb

SLIDE 14

14

Grammar Types

Cont ext sensit ive grammar:

– More t erminals on right -hand side – ASB AAaBB

Recursively enumerable grammar:

– No const raint s

SLIDE 15

15

Lexicon example

Noun breeze | glitter | agent
Verb is | see | smell | shoot
Adj ect ive right | lef t | east | dead
Adverb there | nearby | ahead
Pronoun me | you | I | it
Name John | Mary | Boston
Art icle the | a | an

SLIDE 16

16

Grammar example

S NP VP | S Conjunction S
NP Pronoun | Name | Noun | Article

Noun | NP PP | NP RelClause

VP Verb | VP NP | VP Adjective | VP

PP | VP Adverb

PP
Preposition PP
RelClause
that VP

SLIDE 17

17

Grammat icalit y J udgement s

Grammar Natural language Set of strings

Goal: design grammar to match natural language

agreement

SLIDE 18

18

Grammat icalit y J udgement s

Overgenerat ion examples:

– Me go Bost on. – I smell pit gold wumpus not hing east .

Undergenerat ion example:

– I t hink t he wumpus is smelly

SLIDE 19

19

Synt act ic Analysis

Parsing: process of f inding a parse t ree

f or a given input st ring

I shoot the wumpus pronoun verb proposition noun NP VP NP S

SLIDE 20

20

Top-down parsing

St art wit h S and search f or a t ree t hat

has st ring at leaves

I shoot the wumpus pronoun verb proposition noun NP VP NP S

SLIDE 21

21

Bot t om up parsing

St art wit h st ring and search f or a t ree

t hat has S as root

I shoot the wumpus pronoun verb proposition noun NP VP NP S

SLIDE 22

22

Parsing ef f iciency

Top-down and bot t om up parsing

inef f icient …

– Exponent ial running t ime

Alt ernat ive: chart parsing

– Dynamic progr amming – Cubic running t ime

SLIDE 23

23

Augment ed Grammars

Grammars t end t o overgenerat e

– Ex: “me eat apple”

Augment grammar t o require

– Agreement bet ween subj ect and verb

Ex: “I smells” vs “I smell”

– Agreement bet ween verb subcat egory and complement

Ex: “give t he gold t o me”
Ex: “give me t he gold”

SLIDE 24

24

Parse ambiguit y

Some sent ences have many grammat ical

parses

Example:

– “Fall leaves f all and spring leaves spring”

SLIDE 25

25

Semant ic I nt erpret at ion

Ext ract meaning f rom ut t erances
Tradit ional approach

– Express meaning wit h logic

Problem

– Ambiguous semant ics – Ex: “Helicopt er powered by human f lies”

SLIDE 26

26

Ambiguit y

Possible causes:

– Met onymy: f igure of speech in which one

bj ect is used t o st and f or anot her

– Met aphor: f igure of speech in which a phrase wit h one lit eral meaning is used t o suggest a dif f erent meaning by analogy – Vagueness – Unknown cont ext

SLIDE 27

27

Cont ext / Experience

Meaning of t en grounded in experience
But humans and machines have dif f erent

experiences because of dif f erent sensors…

I s t hat a problem f or nat ural language

underst anding?

SLIDE 28

28

Next Class

Next Class:
Probabilist ic Language Processing
Russell and Norvig Ch. 23