March 1, 2009 Dr. Muhammed Al-Mulhem 1 ICS 482 Natural Language Processing Semantics (Chapter 17) Muhammed Al-Mulhem March 1, 2009

March 1, 2009 Dr. Muhammed Al-Mulhem 1

ICS 482ICS 482Natural Language Natural Language

ProcessingProcessing

Semantics (Chapter 17)Semantics (Chapter 17)

Muhammed Al-MulhemMuhammed Al-Mulhem

March 1, 2009March 1, 2009

March 1, 2009 2Dr. Muhammed Al-Mulhem

MeaningMeaning So far, we have focused on the structure So far, we have focused on the structure

of language – not on the of language – not on the meaning. Words have different meaning, Words have different meaning,

depending on the context in which they depending on the context in which they are used.are used.

We are going to cover : We are going to cover : What is the meaning of a word. What is the meaning of a word. How can we represent the meaning.How can we represent the meaning. What formalisms can be used .What formalisms can be used .


Meaning Meaning RepresentationsRepresentations

We’re going to take the same basic approach to We’re going to take the same basic approach to meaning that we took to syntax and morphologymeaning that we took to syntax and morphology

We’re going to create representations of We’re going to create representations of linguistic inputs that capture the meanings of linguistic inputs that capture the meanings of those inputsthose inputs

But unlike parse trees and the like these But unlike parse trees and the like these representations aren’t primarily descriptions of representations aren’t primarily descriptions of the structure of the inputs…the structure of the inputs…

In most cases, they’re simultaneously In most cases, they’re simultaneously descriptions of the meanings of utterances and descriptions of the meanings of utterances and of some potential state of affairs in some worldof some potential state of affairs in some world


Meaning Meaning RepresentationsRepresentations

Meaning representations are Meaning representations are representations of linguistic inputs that representations of linguistic inputs that capture the meanings of those inputscapture the meanings of those inputs

They permit or facilitate semantic They permit or facilitate semantic processing:processing:

Permit us to reason about their truth (relationship Permit us to reason about their truth (relationship to some world)to some world)

Permit us to answer questions based on their Permit us to answer questions based on their contentcontent

Permit us to perform inference (answer questions Permit us to perform inference (answer questions and determine the truth of things we don’t actually and determine the truth of things we don’t actually know)know)


Common Meaning Common Meaning RepresentationsRepresentations

Sample meaning representation for the sentence:

I have a car First Order Predicate Calculus (FOPC):

x,yHaving(x) Haver(Speaker,x) HadThing(y,x) Car(y)

Semantic Net:

Having

Haver Had-Thing

Speaker Car


Common Meaning Common Meaning RepresentationsRepresentations

Conceptual Dependency Diagram: Conceptual Dependency Diagram: Car Car

Poss-By Poss-By

Speaker Speaker

Frame-based Representations Frame-based Representations

HavingHaver: Speaker

HadThing: Car


Correspondence Between Correspondence Between RepresentationsRepresentations

They all share a common They all share a common foundation:foundation:

meaning representation consists of meaning representation consists of structures composed of sets of symbols.structures composed of sets of symbols.

Symbols include ObjectsObjects properties of objectsproperties of objects relations among objects relations among objects


Two Distinct Two Distinct PerspectivesPerspectives

These four representations can be These four representations can be viewed from two distinct viewed from two distinct perspectives:perspectives:

All represent the meaning of a All represent the meaning of a particular linguistic input particular linguistic input I have a car I have a car

All represent the state of affair in All represent the state of affair in some worldsome world


Requirements of meaning Requirements of meaning representationsrepresentations

Requirements that a meaning representation must fulfill: Verifiability Ambiguity Canonical Form Inference Expressiveness


VerifiabilityVerifiability

The system’s ability to compare the The system’s ability to compare the representation of the meaning of a sentence representation of the meaning of a sentence with the representation in a knowledge base. with the representation in a knowledge base.

ExampleExampleDoes Herfi serve vegetarian food? Does Herfi serve vegetarian food?

Serves (Herfi, vegetarian food) Serves (Herfi, vegetarian food) Match this representation with the Match this representation with the

knowledge base:knowledge base: If found return trueIf found return true Otherwise return false. Otherwise return false.


AmbiguityAmbiguity The system should allow us to The system should allow us to

represent meanings represent meanings unambiguously. unambiguously.


Canonical FormCanonical Form Distinct inputs could have the same Distinct inputs could have the same

meaning meaning Does Herfi serve vegetarian dishes? Does Herfi serve vegetarian dishes? Do they have vegetarian food at Do they have vegetarian food at

Herfi? Herfi? Are vegetarian dishes served at Are vegetarian dishes served at

Herfi? Herfi? Does Herfi serve vegetarian fare? Does Herfi serve vegetarian fare?


Canonical FormCanonical Form Solution: Solution: Inputs that mean the same thing should Inputs that mean the same thing should

have the same meaning representation have the same meaning representation Vegetarian dishes, vegetarian food, Vegetarian dishes, vegetarian food,

vegetarian fare. vegetarian fare. Have, serve Have, serve

Relations among objects to be identicalRelations among objects to be identical syntactic role analysis (e.g., subjects and syntactic role analysis (e.g., subjects and

objects) objects) Herfi serves vegetarian dishes Herfi serves vegetarian dishes Vegetarian dishes are served by Herfi Vegetarian dishes are served by Herfi


InferenceInference Consider a more complex request

Can vegetarians eat at Herfi? It would be a mistake to invoke the canonical

form to force the system to assign the same representation to this request as those of:

Does Herfi serve vegetarian food? Both results in the same answer, not

because they mean the same thing, but because there is a common sense between what vegetarian eats and what vegetarian restaurants serve.


InferenceInference Inference: system’s ability to draw Inference: system’s ability to draw

valid conclusions based on the valid conclusions based on the meaning representation of inputs and meaning representation of inputs and its store of background knowledge. its store of background knowledge.

The system must draw conclusions The system must draw conclusions about the truth of propositions that about the truth of propositions that are not explicitly represented in the are not explicitly represented in the knowledge base, but that are knowledge base, but that are logically derivable from the logically derivable from the propositions that are present.propositions that are present.


Variables for inferenceVariables for inferenceI’d like to find a restaurant where I I’d like to find a restaurant where I

can get vegetarian food can get vegetarian food First observation: First observation:

The request does not make reference The request does not make reference to any particular restaurant to any particular restaurant

Use of variables since we do not know Use of variables since we do not know the name of restaurant the name of restaurant

A representation can be: A representation can be: Serves(X, vegetarianFood) Serves(X, vegetarianFood)


ExpressivenessExpressiveness Must accommodate wide variety of Must accommodate wide variety of

meanings meanings First Order Predicate Calculus First Order Predicate Calculus

(FOPC) is expressive enough to (FOPC) is expressive enough to handle many of the NLP needs. handle many of the NLP needs.

Documents

March 1, 2009 Dr. Muhammed Al-Mulhem 1 ICS 482 Natural Language Processing Semantics (Chapter 17) Muhammed Al-Mulhem March 1, 2009