Wordnet Introduction

WORDNET Reporter: Nguyen Duc Minh Khoi

@ Ho Chi Minh City University of Technology

Thursday, November 01, 2012

Contents

Intro to WordNet

Nouns

Modifiers

Verbs

WordNet system

11/1/2012 WordNet Report 2

INTRODUCTION TO WORDNET


Overview

• WordNet is lexical database for the English language that groups English word into set of synonyms called synset

• Authors: the Cognitive Science Laboratory of Princeton University under the direction of psychology professor George A. Miller

• Used by: • Linguistics Scientist

• Psychologist

• Artificial intelligence Scientist

• Natural Language Processing Scientist


Contents of WordNet

• WordNet distinguish between nouns, verbs, adjectives, adverbs – 4 major syntactic categories

• WordNet contains basic units: • Compounds

• Phrasal verbs

• Collocations

• Idiomatic phrases

• WordNet as a dictionary: • Give definitions

• Sample sentences

• Contains synonym sets

• WordNet as a thesaurus: • Conceptual level: semantic conceptual relations

• Lexical level: lexical relation


Other information

• Lexical database can be built by: • Automatic acquisition

• Craft one dictionary by hand

• Knowledge engineering: • Lexical level: contains information about synonyms, antonyms...

• Domain level: refer to the topic of discourse

• Application specific level: relates objects and events

• Tennis problem: • Contains no relations that indicate the word’s shared membership in a

topic of discourse

• E.g. not link racquet, ball, net => court game


NOUNS


Introduction to nouns in WordNet

• WordNet is machine readable dictionary

• Noun in WordNet doesn’t give: • pronunciation

• Derivative morphology

• Etymology

• Usage notes

• Pictorial illustration

• WordNet try to make semantic relations by extract synonym from thesaurus manually

• WordNet lexicalized concept by making synset relate to that concept


Lexical hierarchy

• Tree graph: graph without circular loop

• Assumptions: • Longer distance in hierarchy longer traverse in thoughts

• More lexical information must be stored in every lexicalized concepts than is required to establish in hierarchy.

• Noun’s unique beginner:


Lexical hierarchy (cont.)

• Examples:


Noun relations • Hyponyms (~):

• A word of more specific meaning than a general or superordinate term applicable to it.

• For example, {bowl} is a hyponym of {dish}: {bowl} ~-> {dish}

• Hypernyms (@): • A word with a broad meaning that more specific words fall under; a

superordinate.

• For example, {color} is a hypernym of {red}: {color} @-> {red}

• Meronyms (#): • The semantic relation that holds between a part and the whole.

• For example, {beak} and {wing} are meronyms of {bird}: {beak, wing} #-> bird

• Three kinds: component, member, made from

• Holonyms (%): • The semantic relation that holds between a whole and its parts

• For example, {building} is a holonym of {window}: {building} %-> {window}


Noun relations (cont.)

• Antonyms (!): • A word opposite in meaning to another

• For example, {man} !-> {woman}

• Polysemous nouns: • Nous that have many meanings

• For example, {mouse} living animal or computer device

• Rules: two meanings of a word are similar then the meaning of their hyponyms should also be similar in the same way.

• Attribute (=) and modifications: • Values of attribute are expressed by adjectives

• Modification can also be nouns

• For examples, chair -> small chair, big chair


MODIFIERS


Adjectives

• Main functions: modifying nouns

• Types: • Descriptive adjectives

• Participle adjectives

• Relational adjectives

• Format: • A(x) = adj

• E.g.: WEIGHT(package) = heavy


Adjectives Relations • Antonyms (!):

• Basic semantic relation among descriptive adjectives

• Means “IS ANOYNYMOUS TO”, e.g. heavy is anonymous to light

• Can be direct, e.g. heavy/light

• Or can be indirect, e.g. heavy/airy


Adjectives Relations (cont.)

• Other relations • Troponym (~):

• Hypernym (@):

• Entailment (*):

• Cause (>):

• Also see (^):


Gradation

• Contrary: one of propositions can be true or both are false

• Adjectives can be use to express different level of action

• For example:


Other stuffs

• Markedness: • Normal linguistic unit (unmarked term) compare to unit possible

irregular forms (marked term)

• E.g.: The pool is 5 feet deep, NOT: The pool is 5 feet shallow

• So deep marked term, shallow unmarked term

• Polysemy and selectional preferences: • E.g.: old can be not young modify persons

old can be not new modify things

• Some adjectives can modify almost any nouns

• E.g.: good / bad, desirable / undesirable

• Some adjectives can strictly restricted to some nouns

• E.g.: editable / ineditable


Other types of descriptive adjectives

• Color adjectives: • Server as nouns and adjectives

• Quantifiers: • E.g.: all, some, many, few…

• Participle adjectives: • Means “PRINCIPLE PART OF”

• E.g.: breaking is principle part of break

• Can be –ing/-ed: running water, elapsed time


Relational adjectives

• Differ from descriptive adjectives by • Do not relate to attribute of nouns

• Can not be gradable

• Occur only attribute position

• Lack of direct antonym

• E.g.: criminal behavior


Adverbs

• Derived from adjectives by suffixation: • -ly:

• Specify manner: e.g.: beautifully

• Specify degree: e.g.: extremely

• Other suffix:

• -wise, -way, -ward

• E.g.: northward, forward

• Inherit their adjectives about: • Antonym

• Gradation


VERBS


Organizations

• Types of semantic verbs: • motion, perception, communication, competition, change, cognitive,

consumption, creation, emotion, possession, body care, functions, social behavior, interaction.

• Stative verb: • Collaborate with be: resemble, belong, suffice

• Control verb: want, fail, prevent, succeed, begin

• Cannot group all verbs in unique beginner like nouns

• English has fewer verb than nouns BUT approximate twice as polysemous as noun

• Verb synset: • Synonym and near synonym: e.g.: pass away vs. die vs. kick the bucket

• Idiom and metaphors: • Kick the bucket include synset

• Die include synonym: break, break down (for car and computer)


Verb Relations

• Entailment (*): • The verb Y is entailed by X if by doing X you must be doing Y

• E.g.: to snore entails to sleep

• Not mutual: V1 * V2 NOT V2 V1


Verb relations

• Troponym (~): • The verb Y is a troponym of the verb X if the activity Y is doing X in

some manner

• E.g.: to lisp is a troponym of to talk

• Special case of entailment

• Most frequently coded in WordNet

• Antonym (!): • E.g.: give/take, buy/sell, lend/borrow, teach/learn

• Can also be troponym: fail/succeed entails try, forget entails know

• Hypernym (@): • The verb Y is a hypernym of the verb X if the activity X is a (kind of) Y

• E.g.: to perceive is an hypernym of to listen


WORDNET SYSTEM


WordNet system


Lexical files

• WordNet store nouns, adjectives, adverbs and nouns into synset lexical source files by syntactic categories • Nouns and verbs: grouped according to semantic fields

• Adjectives are divided among three files (adj.all, adj.ppl, adj.pert)

• Adverb are store in single file

• Relation pointers store in WordNet


Sample Application use WordNet

• NLTK is a platform for building Python programs to work with human language data

• Sample commands: • Work with nouns:


Sample Application use WordNet (cont.)


• Work with verbs