MethodsforIdentifyingRelationsandAssociationsinText

StuartG.Townsand RichardWatsonToddKingMongkut’sUniversityofTechnologyThonburi

DRAL2017

PreviousResearchintoWritingQuality

Whatarethelinguisticfeaturesfoundinproficientwriting?

Whatmakesgoodwritinggood?

Howdoweteachstudentstobegoodwriters?

Howarelinguisticfeaturesexpressedindifferentlevelsofproficiency?

Whatlinguisticfeaturescorrelatewithhumanjudgementsofwritingquality?

WhatisBasicEnglish?

WhatdoesEnglishProficiencymean?

LinguisticFeaturesforWritingQuality

SyntacticComplexity• ClauseLength• #ofwordsbeforethemainverb

LexicalDiversity&Sophistication• Type-TokenRatios• WordRarity

Cohesion• Repetition&Reference• Paraphrases

Coherence

?

ResearchonWritingQuality

Pulitzer Prize Winner

LinguisticFeaturesforWritingQuality

SyntacticComplexity• ClauseLength• #ofwordsbeforethemainverb

LexicalDiversity&Sophistication• Type-TokenRatios• WordRarity

Cohesion• Repetition&Reference• Paraphrases

Coherence

?

ConnectednessofConceptsinaText

Cohesion Coherenceexplicitconnectedconceptsinthetext

implicitconnectionsinthemindofthereader

Reiterations• Repetition• Reference

Relations• Synonyms• Hyponyms• Meronyms

Associations• Anyother

connection

“Moonrise Kingdom” opens with no music — just the sound of raindrops falling on the roof of apreternaturally cozy house, which the camera gently leads the audience through as the familymembers inside go about their rainy day business.

Bathed in apple reds, egg yolk yellows and an air of studied eccentricity, the house is immediatelyrecognizable as yet another habitat created by Wes Anderson, a film director whose obsession withmaterial culture, nostalgia and nursery comforts borders on the fetishistic.

Of course, for viewers who happen to share Anderson’s taste for boldly framed, bespokeproductions — in which everything looks (and most probably is) lovingly handmade and artisanal,“Moonrise Kingdom” will simply offer yet another chance to live, at least for a little while, in thekind of universe only Anderson can create.

(You can almost smell the damp canvas and wood polish in that opening sequence.)

Those who long ago wrote off the writer-director as insufferably mannered and arcane — the usualterm of art is “twee” — well, they’re welcome to stay out in the rain.

That opening scene house has a name, by the way: Summer’s End, which turns out to aptly capturea vaguely autumnal tale of young love that takes place in early September 1965 — a time of FordFalcons and mothers who smoked.

“Moonrise Kingdom” opens with no music — just the sound of raindrops falling on the roof of apreternaturally cozy house, which the camera gently leads the audience through as the familymembers inside go about their rainy day business.

Bathed in apple reds, egg yolk yellows and an air of studied eccentricity, the house is immediatelyrecognizable as yet another habitat created by Wes Anderson, a film director whose obsession withmaterial culture, nostalgia and nursery comforts borders on the fetishistic.

Of course, for viewers who happen to share Anderson’s taste for boldly framed, bespokeproductions — in which everything looks (and most probably is) lovingly handmade and artisanal,“Moonrise Kingdom” will simply offer yet another chance to live, at least for a little while, in thekind of universe only Anderson can create.

(You can almost smell the damp canvas and wood polish in that opening sequence.)

Those who long ago wrote off the writer-director as insufferably mannered and arcane — the usualterm of art is “twee” — well, they’re welcome to stay out in the rain.

That opening scene house has a name, by the way: Summer’s End, which turns out to aptly capturea vaguely autumnal tale of young love that takes place in early September 1965 — a time of FordFalcons and mothers who smoked.

“Moonrise Kingdom” opens with no music — just the sound of raindrops falling on the roof of apreternaturally cozy house, which the camera gently leads the audience through as the familymembers inside go about their rainy day business.

Bathed in apple reds, egg yolk yellows and an air of studied eccentricity, the house is immediatelyrecognizable as yet another habitat created by Wes Anderson, a film director whose obsessionwith material culture, nostalgia and nursery comforts borders on the fetishistic.

Of course, for viewers who happen to share Anderson’s taste for boldly framed, bespokeproductions — in which everything looks (and most probably is) lovingly handmade and artisanal,“Moonrise Kingdom” will simply offer yet another chance to live, at least for a little while, in thekind of universe only Anderson can create.

(You can almost smell the damp canvas and wood polish in that opening sequence.)

Those who long ago wrote off the writer-director as insufferably mannered and arcane — the usualterm of art is “twee” — well, they’re welcome to stay out in the rain.

That opening scene house has a name, by the way: Summer’s End, which turns out to aptly capturea vaguely autumnal tale of young love that takes place in early September 1965 — a time of FordFalcons andmothers who smoked.

“Moonrise Kingdom” opens with no music — just the sound of raindrops falling on the roof of apreternaturally cozy house, which the camera gently leads the audience through as the familymembers inside go about their rainy day business.

Bathed in apple reds, egg yolk yellows and an air of studied eccentricity, the house is immediatelyrecognizable as yet another habitat created by Wes Anderson, a film director whose obsessionwith material culture, nostalgia and nursery comforts borders on the fetishistic.

Of course, for viewers who happen to share Anderson’s taste for boldly framed, bespokeproductions — in which everything looks (and most probably is) lovingly handmade and artisanal,“Moonrise Kingdom” will simply offer yet another chance to live, at least for a little while, in thekind of universe only Anderson can create.

(You can almost smell the damp canvas and wood polish in that opening sequence.)

Those who long ago wrote off the writer-director as insufferably mannered and arcane — theusual term of art is “twee” — well, they’re welcome to stay out in the rain.

That opening scene house has a name, by the way: Summer’s End, which turns out to aptlycapture a vaguely autumnal tale of young love that takes place in early September 1965 — a timeof Ford Falcons and mothers who smoked.

“Moonrise Kingdom” opens with no music — just the sound of raindrops falling on the roof of apreternaturally cozy house, which the camera gently leads the audience through as the familymembers inside go about their rainy day business.

Bathed in apple reds, egg yolk yellows and an air of studied eccentricity, the house is immediatelyrecognizable as yet another habitat created by Wes Anderson, a film director whose obsessionwith material culture, nostalgia and nursery comforts borders on the fetishistic.

Of course, for viewers who happen to share Anderson’s taste for boldly framed, bespokeproductions — in which everything looks (and most probably is) lovingly handmade and artisanal,“Moonrise Kingdom” will simply offer yet another chance to live, at least for a little while, in thekind of universe only Anderson can create.

(You can almost smell the damp canvas and wood polish in that opening sequence.)

Those who long ago wrote off the writer-director as insufferably mannered and arcane — theusual term of art is “twee” — well, they’re welcome to stay out in the rain.

That opening scene house has a name, by the way: Summer’s End, which turns out to aptlycapture a vaguely autumnal tale of young love that takes place in early September 1965 — a timeof Ford Falcons andmothers who smoked.

Associations:PPWsvsBloggers

Author BlogorPPW? %Assoc.

MO PPW 56%

DA PPW 40%

HD PPW 38%

MG PPW 31%

AF Blog 26%

FB Blog 23%

FI Blog 16%

CN Blog 15%

Author BlogorPPW? %Assoc.

HD PPW 43%

DA PPW 41%

MO PPW 41%

MG PPW 32%

CT Blog 30%

CN Blog 30%

FI Blog 24%

AF Blog 22%

MoonriseKingdom GhostProtocol

MethodsforIdentifyingRelationsandAssociationsinText

Relations:• OxfordAmericanWriter'sThesaurus• WordNet(Princeton)• UCRELSemanticAnalysisSystem

Associations:• WordassociationdatabaseatSmallWorldofWords• NearNeighborsLSAtoolhostedattheUniversityofColorado• MIscoresfromtheCorpusofContemporaryAmericanEnglish(COCA)

CreatingaWordList

Thesecriteriawerefollowedtoidentify73wordstobeanalyzed:

• Mustbecontentwordsonly,excludingadverbs,propernouns,auxiliaryverbs,phrasalverbs,orotheridioms(e.g.ofcourse,bytheway)

• Mustbeseparateentriesindictionariesandthesauruses(e.g.rainy-daywas brokenupintoseparatewordsbutegg-yolkwasnot)

• Mustnotbeoneofthetop250mostcommonEnglishwords

ThreePhasesofthisStudy

• Phase1: Identifyallpossiblewordpairconnectionspersource,thendeterminewhichwordpairconnectionsarerelevanttoourtext.

• Phase2:Determinewhichofthesixsourcesreturnedthemostvalidresults.

• Phase3:Sumtheresultsfromallsixsourcestoseeifthewordpairsmatchedtheresearcher’sintuitionofconnectednessbetweenconceptsinthetext.

Phase1

Example:Findingrelevantconnections

IntheThesaurus:Thehouseistheaudienceatatheatre

Inourtext:Theaudienceiswatchingthemoviethatshowsahouse

audience– house

Phase1

Identifyallpossiblewordpairconnectionspersource,thendeterminewhichwordpairconnectionsarerelevanttoourtext.

Source Identified Relevant %RelevantThesaurus 21 6 29%

WordNet 32 12 38%

Semantic Tagging 21 10 48%

SmallWorld 45 23 51%

LSA 34 19 56%

MIScores 28 15 54%

(Note:73words->2,628possiblecombinations)

Phase2

Determinewhichofthesixsourcesreturnedthemostvalidresults.

• Assumption:Anexceptionallywell-writtentextshouldhavehigh-connectednessandshouldbewell-organized.I.e.,theconnectedpairsshouldbeclosertogetherthanarandomorder.

• PointBi-SerialCorrelationbetweendistancebetweenthewordsinawordpairandwhetherornotitwasamatch.

Phase2

*p<.05**p<.00001

Source PBSCorrelationThesaurus -0.17**

WordNet -0.04*

SemanticTagging -0.10**

SmallWorldofWords -0.24**

LatentSemanticAnalysis -0.02

COCAMIScores -0.12**

PointBiserialCorrelationbetweendistancesandrelevance

Phase3

Sumtheresultsfromallsixsourcestoseeifthewordpairsmatchedtheresearcher’sintuitionofconnectednessbetweenconcepts.

audience– viewersreds– yellowsdamp- rain

Foundby… #Word Pairs %Relevant6sources 0 n/a

5Sources 0 n/a

4Sources 3 100%

3Sources 8 88%

2Sources 27 52%

1Source 91 29%

0Sources 2,499 0%

Phase1FindingSummary

Phase1:Theassociationdatabase(SmallWorld)andcorpus-basedsources(LSAandMI)returnedbetterresultsthantheothersources,asitidentified:

• Morewordpairs

• Higherpercentageofrelevantwordpairs

• Uniquereiterations,relations,andassociations

Phase2FindingSummary

• TheSmallWorldwasthemostvalidsource(withrespecttodistancebetweenwords),withWordNetandLSAbeingtheleastvalid.

However,WordNetandLSAidentifiedthemostnumberofmovieterms.

• Limitation:Thisstudyonlyconsideredwordpairs,notwordchainsorwordnetworks.

Phase3FindingSummary

Themoreoftenawordpairwasfound,themorelikelyitwasarelevantwordpairforourtext.

Thesemethodsreducedthenumberofpotentialwordpairsfrom2,628toamoremanageable129.

Thesesixmethodsthereforeshowpromiseforfuturefullautomationoftheidentificationofrelationsandassociationsinatext.

Thankyou!