cs626-460: language technology for the web/natural language processing pushpak bhattacharyya cse...

25
CS626-460: Language CS626-460: Language Technology for the Technology for the Web/Natural Language Web/Natural Language Processing Processing Pushpak Bhattacharyya Pushpak Bhattacharyya CSE Dept., CSE Dept., IIT Bombay IIT Bombay Topic: More on semantic Topic: More on semantic relations relations

Upload: june-elliott

Post on 26-Dec-2015

218 views

Category:

Documents


1 download

TRANSCRIPT

CS626-460: Language CS626-460: Language Technology for the Web/Natural Technology for the Web/Natural

Language ProcessingLanguage Processing

Pushpak BhattacharyyaPushpak BhattacharyyaCSE Dept., CSE Dept., IIT Bombay IIT Bombay

Topic: More on semantic relationsTopic: More on semantic relations

MetonymyMetonymy

• Associated with Associated with Metaphors Metaphors which are which are epitomes of semanticsepitomes of semantics

• Oxford Advanced Learners Dictionary Oxford Advanced Learners Dictionary definition: “The use of a word or definition: “The use of a word or phrase to mean something different phrase to mean something different from the literal meaning”from the literal meaning”

• Does it mean Careless Usage?!Does it mean Careless Usage?!

Insight from Sanskritic Insight from Sanskritic TraditionTradition

• Power of a wordPower of a word– Abhidha, Lakshana, VyanjanaAbhidha, Lakshana, Vyanjana

• Meaning of Meaning of Hall:Hall:– The hall is packed (avidha)The hall is packed (avidha)– The hall burst into laughing (lakshana)The hall burst into laughing (lakshana)– The Hall is full (The Hall is full (unsaidunsaid: and so we : and so we

cannot enter) (vyanjana)cannot enter) (vyanjana)

Metaphors in Indian Metaphors in Indian TraditionTradition

• upamana upamana and and upameya upameya – Former: object being comparedFormer: object being compared– Latter: object being compared withLatter: object being compared with– Puru was like a lion in the battle with Puru was like a lion in the battle with

Alexander Alexander (Puru: (Puru: upameya; upameya; Lion: Lion: upamana)upamana)

Upamana, rupak, Upamana, rupak, atishayoktiatishayokti• upamanaupamana: Explicit comparison: Explicit comparison

– Puru was like a lion in the battle with Puru was like a lion in the battle with AlexanderAlexander

• rupakrupak: Implicit comparison: Implicit comparison– Puru was a lion in the battle with Puru was a lion in the battle with

AlexanderAlexander

• Atishayokti (exaggeration)Atishayokti (exaggeration): upamana : upamana and upameya droppedand upameya dropped– Puru’s army fled. But the lion fought on.Puru’s army fled. But the lion fought on.

Modern study (1956 onwards, Modern study (1956 onwards, Richards et. al.)Richards et. al.)• Three constituents of metaphorThree constituents of metaphor

– Vehicle Vehicle (items used metaphorically)(items used metaphorically)– Tenor Tenor (the metaphorical meaning of the (the metaphorical meaning of the

former)former)– Ground Ground (the basis for metaphorical extension)(the basis for metaphorical extension)

• ““The foot of the mountain”The foot of the mountain” – Vehicle: :foot”Vehicle: :foot”– Tenor: “lower portion”Tenor: “lower portion”– Ground: “spatial parallel between the Ground: “spatial parallel between the

relationship between the foot to the human relationship between the foot to the human body and the lower portion of the mountain body and the lower portion of the mountain with the rest of the mountain”with the rest of the mountain”

Interaction of semantic Interaction of semantic fieldsfields((Haas)Haas)• Core vs. peripheral semantic fieldsCore vs. peripheral semantic fields

• Interaction of two words in Interaction of two words in metonymic relation brings in new metonymic relation brings in new semantic fields with selective semantic fields with selective inclusion of featuresinclusion of features

• Leg of a tableLeg of a table– Does not Does not stretchstretch or or movemove– Does Does stand stand and and supportsupport

Lakoff’s (1987) contributionLakoff’s (1987) contribution

• Source DomainSource Domain

• Target DomainTarget Domain

• Mapping RelationsMapping Relations

Mapping Relations: ontological Mapping Relations: ontological correspondencescorrespondences

• Anger is heat of Anger is heat of fluid in containerfluid in container

HeatHeat

(i) Container(i) Container

(ii) Agitation (ii) Agitation of fluidof fluid

(iii) Limit of (iii) Limit of resistenceresistence

(iv) Explosion(iv) Explosion

AngerAnger

BodyBody

Agitation of Agitation of mindmind

Limit of ability Limit of ability to suppressto suppress

Loss of Loss of controlcontrol

Image SchemasImage Schemas

• Categories: Container ContainedCategories: Container Contained• QuantityQuantity

– More is up, less is down: More is up, less is down: Outputs rose Outputs rose dramatically; accidents rates were lowerdramatically; accidents rates were lower

– Linear scales and paths: Linear scales and paths: Ram is by far the best Ram is by far the best performerperformer

• TimeTime– Stationary event: Stationary event: we are coming to exam timewe are coming to exam time– Stationary observer: Stationary observer: weeks rush byweeks rush by

• Causation: Causation: desperation drove her to desperation drove her to exterme stepsexterme steps

Patterns of MetonymyPatterns of Metonymy

• Container for containedContainer for contained– The kettle boiled The kettle boiled (water)(water)

• Possessor for possessed/attributePossessor for possessed/attribute– Where are you parked? Where are you parked? (car)(car)

• Represented entity for representativeRepresented entity for representative– The government will announce new The government will announce new

targetstargets

• Whole for partWhole for part– I am going fill up the car with petrolI am going fill up the car with petrol

Patterns of Metonymy Patterns of Metonymy (contd)(contd)

• Part for wholePart for whole– I noticed several new faces in the classI noticed several new faces in the class

• Place for institutionPlace for institution– Lalbaug witnessed the largest Ganapati Lalbaug witnessed the largest Ganapati

Question: Can you have part-part Question: Can you have part-part metonymy metonymy

Purpose of MetonymyPurpose of Metonymy

• More idiomatic/natural way of expressionMore idiomatic/natural way of expression– More natural to say More natural to say the kettle is boiling the kettle is boiling as as

opposed to opposed to the water in the kettle is boilingthe water in the kettle is boiling• EconomyEconomy

– Room 23 is answering Room 23 is answering (but not *(but not *is asleepis asleep))• Ease of access to referentEase of access to referent

– He is in the phone book He is in the phone book (but not *on the back (but not *on the back of my hand)of my hand)

• Highlighting of associated relationHighlighting of associated relation– The car in the front decided to turn rightThe car in the front decided to turn right (but (but

not *not *to smoke a cigarette)to smoke a cigarette)

Feature sharing not Feature sharing not necessarynecessary

• In a restaurant:In a restaurant:– Jalebii ko abhi dudh chaiye Jalebii ko abhi dudh chaiye (no feature (no feature

sharing)sharing)– The elephant now want some coffee The elephant now want some coffee

(feature sharing)(feature sharing)

ProverbsProverbs

• Describes a specific event or state of Describes a specific event or state of affairs which is applicable affairs which is applicable metaphorically to a range of events metaphorically to a range of events or states of affairs provided they or states of affairs provided they have the same of sufficiently similar have the same of sufficiently similar image-schematic structureimage-schematic structure

Formalization of WNFormalization of WN

Formalization of WordNetFormalization of WordNet

•Mathematical Formal TheoryMathematical Formal Theory

•Based on Formal Concept Based on Formal Concept AnalysisAnalysis

•Defines Semantic RelationshipsDefines Semantic Relationships

•Theoretical Analysis of Theoretical Analysis of Relationships and dependencies Relationships and dependencies among themamong them

Formal Concept AnalysisFormal Concept Analysis

• Formal Context K : (G,M,I)Formal Context K : (G,M,I)

• G : Set of Formal ObjectsG : Set of Formal Objects

• M : Set of Formal AttributesM : Set of Formal Attributes

• I G MI G M

• A pair (A,B) is a Formal Concept if,A pair (A,B) is a Formal Concept if,

A = A = ββB and B = B and B = ααAA

Example of Formal ConceptExample of Formal ConceptYoungYoung OldOld FemaleFemale MaleMale

PersonPerson

AdultAdult XX

Female PersonFemale Person XX

Male PersonMale Person XX

ChildChild XX

WomanWoman XX XX

ManMan XX XX

A = {Adult, Woman, Man} & B = {Old}A = {Adult, Woman, Man} & B = {Old}Here (A,B) is a Formal Concept. Here (A,B) is a Formal Concept.

WordNet as a Formal WordNet as a Formal ContextContext• Denotative Context :Denotative Context :

here D is Denotata & concepts are here D is Denotata & concepts are Disambiguated WordsDisambiguated Words

• WordNet is a Lexical Context: WordNet is a Lexical Context: here W are here W are Disambiguated WordsDisambiguated Words

• An equivalence relation SYN is defined on An equivalence relation SYN is defined on set W as, set W as, where syn(w) is Synset of wwhere syn(w) is Synset of w

Relational Concept AnalysisRelational Concept Analysis• Relations between concepts, i.e.Relations between concepts, i.e.

• Quantifiers are used to define such Quantifiers are used to define such relationshipsrelationshipsas follows:as follows:

• Notations :Notations :

MeronymyMeronymy

• Two disambiguated words are in Two disambiguated words are in Meronymy Relation if their denotative Meronymy Relation if their denotative word concepts are in relation word concepts are in relation where m is meronymy relation among where m is meronymy relation among denotata i.e.denotata i.e.

• Various combinations of quantifiers lead to Various combinations of quantifiers lead to different types of Meronymy Relationsdifferent types of Meronymy Relations

Types of MeronymyTypes of Meronymy

• : A child can be a member of tennis club, : A child can be a member of tennis club, but not all children are members of tennis but not all children are members of tennis club, nor do all tennis clubs have children as club, nor do all tennis clubs have children as members. members.

• : All door-handles are part of doors, : All door-handles are part of doors, but not all doors have to have handles.but not all doors have to have handles.

• : All sandwitches consist of breads, : All sandwitches consist of breads, but not all breads are parts of sandwitches.but not all breads are parts of sandwitches.

• : Each bird feather is part of the bird : Each bird feather is part of the bird and all birds have feathers.and all birds have feathers.

Hyponymy and SynonymyHyponymy and Synonymy

• A disambiguated word is a Hyponym of A disambiguated word is a Hyponym of other word if the concept it denotes is other word if the concept it denotes is subconcept of concept denoted by the subconcept of concept denoted by the other word.other word.

• Two disambiguated words are called Two disambiguated words are called Synonyms if they denote the same Synonyms if they denote the same concept.concept.

Inheritence of RelationsInheritence of Relations

• : This is inherited by hypernyms of : This is inherited by hypernyms of part and whole.part and whole.

• : This is inherited by hypernyms : This is inherited by hypernyms of whole and hyponyms of part.of whole and hyponyms of part.

• : This is inherited by hyponyms : This is inherited by hyponyms of whole and hypernyms of part.of whole and hypernyms of part.

• : This is inherited by a : This is inherited by a particular pair hyponyms of both.particular pair hyponyms of both.