towards distributed grammar motivations and issues of representation andré wlodarczyk

51
Towards Distributed Towards Distributed Grammar Grammar motivations and issues of motivations and issues of representation representation André WLODARCZYK André WLODARCZYK

Upload: margery-obrien

Post on 16-Jan-2016

217 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Towards Distributed GrammarTowards Distributed Grammarmotivations and issues of representationmotivations and issues of representation

Towards Distributed GrammarTowards Distributed Grammarmotivations and issues of representationmotivations and issues of representation

André WLODARCZYKAndré WLODARCZYK

Page 2: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Preliminary remarksPreliminary remarksPreliminary remarksPreliminary remarks

Page 3: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Characteristics of Human LanguagesCharacteristics of Human LanguagesCharacteristics of Human LanguagesCharacteristics of Human Languages

1. Ambiguity (prototype and open-endedness)2. Dynamics/Emergence (various changes)3. Partiality (expressions need grounding and refinement)4. Learnability (language must be learned)5. Composability (only partial composition)6. Recursivity (only constrained recursion)7. Distributivity (sequential and parallel processes)8. Complexity (hard tractability)9. Typicality (langue uses can be typed)etc…

Page 4: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

The Meaning of a SentenceThe Meaning of a Sentencein a mono-dimensional development hypothesisin a mono-dimensional development hypothesis

The Meaning of a SentenceThe Meaning of a Sentencein a mono-dimensional development hypothesisin a mono-dimensional development hypothesis

Knowledge

Legend:The Surface Structure of a sentence s is transformed into the Deep Structure tree representation α.

α = M(s)

LogicalForm

In Generative Semanticsα stands for the meaning of a sentence s as transformed into a Logical Form. Deep Structure

Surface Structure

Page 5: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

The Contents of DiscourseThe Contents of Discoursebi-dimensional development hypothesisbi-dimensional development hypothesis

Salience

Information

Salience

Information

Relevance

Knowledge

SemanticContents S(Φ)

PragmaticContents P(Φ)

DownwardDevelopment

UpwardDevelopment

Discourse level

Legend:S(Φ) stands for a set of information-related formulaeP(Φ) stands for a set of discourse-related formulae

Page 6: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Main Dimensions of CognitionMain Dimensions of Cognition

Attention: Salience gives rise to the selective centering of information.

Intention: “Relevance relies on intentional attitudes which are guided by the attentional control of Salience” (I. Kecskes).

Emotion: Emergence and Deconstruction are determined as well by the attentional centering of information as by the intentionally guided Relevance of Knowledge.

Relevance

Knowledge Information

SalienceIntentionAttention

Deconstruction

EmergenceEmotion

Page 7: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Main Degrees of DevelopmentMain Degrees of Development

Standard Standard meaningmeaning

Concise Concise (underdeveloped) (underdeveloped) meaningmeaning

Precise Precise (developed) (developed) meaningmeaning

Page 8: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

AS & MIC AS & MIC AS & MIC AS & MIC CONTENTCONTENT

SemanticsSemantics PragmaticsPragmatics

Bi-axial ordering Bi-axial ordering of discourseof discourse

SignificationSignification

(Information)(Information)CommunicationCommunication

(Meta-Information)(Meta-Information)

SelectionSelection

(paradigmatic)(paradigmatic)

Property Property ComparisonComparison

(x X)∈(x X)∈

(Attribute Space)(Attribute Space)

CenteringCentering

(distinguish x)(distinguish x)

CombinationCombination

(syntagmatic)(syntagmatic)

Scenario CreationScenario Creation

r(x)r(x)

(Relation on x)(Relation on x)

PredicationPredication

(say p about x)(say p about x)

AS

AssociativeSemantics

MIC

Meta-InformativeCentering

Page 9: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Information Flow LogicInformation Flow Logicforfor

DISTRIBUTED SYSTEMSDISTRIBUTED SYSTEMS

Page 10: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

LINGUISTIC RESOURCES

Functional System

Distributed Grammar ArchitectureDistributed Grammar Architecture

Sounds,Phonemes

Grammaticaland Lexical Morphemes

DerivationsPhraseology (templates and idioms)

Page 11: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

IF : RefinementIF : RefinementIF : RefinementIF : Refinement

A + B

“INFORMATION FLOW- the Logic of Distributed Systemsthe Logic of Distributed Systems” by Jon BARWISE & Jerry SELIGMAN Cambridge Univerity Press (1997), p. 44.

KK

f’f’ g’g’

CC

AA BB

ff gg

Refinement

Page 12: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Refinement in Distributed GrammarRefinement in Distributed GrammarRefinement in Distributed GrammarRefinement in Distributed Grammar

refinement

Codes

A. Wlodarczyk (2008)

INFONS

NOEMAS

KNOWLEDGEKNOWLEDGE

INFORMATIONINFORMATION

LANGUAGELANGUAGE WORLDWORLD

carryingcarryingconveyingconveying

ThingsThings

Page 13: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

IF : Distributed SystemIF : Distributed SystemIF : Distributed SystemIF : Distributed System

“INFORMATION FLOW- the Logic of Distributed Systemsthe Logic of Distributed Systems” by Jon BARWISE & Jerry SELIGMAN Cambridge Univerity Press (1997), p. 90.

f2f2

g1g1 g2g2

A1 carries information on A3 by A2A1 carries information on A3 by A2

f1f1 f3f3 f4f4

A2A2A1A1 A3A3

CC

B1B1 B2B2

Page 14: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Distributed GrammarDistributed GrammarDistributed GrammarDistributed Grammar

f2f2

g1g1 g2g2

f1f1 f3f3 f4f4

syntaxsyntaxsemanticssemantics pragmaticspragmatics

sensesense

informationinformation meta-informationmeta-information

Page 15: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Distributed Communication SettingDistributed Communication SettingDistributed Communication SettingDistributed Communication Setting

f2f2

g1g1 g2g2

f1f1 f3f3 f4f4

utteranceutterancespeakerspeaker hearerhearer

sensesense

speakingspeaking listeninglistening

Page 16: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

ASSOCIATIVE SEMANTICSASSOCIATIVE SEMANTICS

Page 17: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Information in AS TheoryInformation in AS TheoryInformation in AS TheoryInformation in AS Theory

In Associative Semantics (AS), the kernel information is In Associative Semantics (AS), the kernel information is determined by semantic situations.determined by semantic situations.

DefinitionDefinition. An individualised situation is represented by the . An individualised situation is represented by the three following components: three following components:

a set of a set of (static or dynamic) (static or dynamic) framesframesa set of roles a set of roles (enacted by agents and/or figures)(enacted by agents and/or figures)a set of anchors a set of anchors (indicators of spatio-temporal relations)(indicators of spatio-temporal relations). .

Page 18: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Information in AS TheoryInformation in AS TheoryInformation in AS TheoryInformation in AS Theory

FRAME ROLE ANCHOR

INNER VIEW Analysis SelectionConcurrent

Synchronisation

OUTER VIEW Control ConfigurationDistributed

Synchronisation

Each situation component has two views:

(a) an inner view (view from inside)(b) an outer view (view from outside)

Page 19: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Situation, their Participants and AnchorsSituation, their Participants and AnchorsSituation, their Participants and AnchorsSituation, their Participants and Anchors

1. Situations (Facts of the World)1. Frames

1. States2. Actions

1. Events2. Processes

2. Roles1. Active2. Median3. Passive

3. Anchors1. Spatial locators2. Temporal locators

2. Participants (Entities of the World)1. Agents (Animate Entities)

1. Human2. Non-Human (animal)

2. Figures (Non Animate Entities)1. Material2. Immaterial

3. Anchors (Space and Time Locators)1. Spatial Locators

1. Start2. Path3. Arrival

2. Temporal Locators1. Beiginning2. Course3. End

Page 20: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Situation Example (1a)Situation Example (1a)Situation Example (1a)Situation Example (1a)

Brutus killed Caesar. kill(Brutus, Caesar)

{kill relprop=asymetric :

{Brutus participant=agent : } {Caesar participant=agent : }

{F prop=asym effect=death : }

{ActiveRole type=init : } {PassiveRole type=term : } {ActiveRole > F > PassiveRole : }

{Brutus > F > Caesar} }

Page 21: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Situation Example (1b)Situation Example (1b)Situation Example (1b)Situation Example (1b)Brutus killed Caesar with a knife.

kill(Brutus, Caesar, knife)

[SIT0: ‘kill’

HEAD kill x y z

BODY

{kill relation property=asymetric effect=death: {brutus participant=agent: } {caesar participant=agent: } {knife participant=figure: } {Role1 x=killer generic=active type=initiator } {Role2 y=kilee generic=passive type=terminator } {Role3 y=weapon generic=median type=origin }

[SIT1: HEAD use x y {use prop=asym purpose=tool : {Role1 x=user generic=active type=initiator : } {Role2 y=arm generic=passive type=terminator :} {Role1 < use > Role2 }} FOOT {brutus < use > knife}]

[SIT2: HEAD causeDie x y {causeDie prop=asym effect=death : {Role1 generic=q-active type=origin object=weapon : } {Role2 generic=passive type=terminator } {knife < causeDie > caesar }} FOOT {knife < causeDie > caesar : }]

FOOT {Brutus < use > knife }

{knife < causeDie > Caesar}]

Page 22: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Distributed Semantic AssociativityDistributed Semantic AssociativityDistributed Semantic AssociativityDistributed Semantic Associativity

f2f2

g1g1 g2g2

f1f1 f3f3 f4f4

knifeknifeBrutusBrutus CaesarCaesar

killkill

useuse cause deathcause death

Example: Brutus stabbed Caesar with a knife.

Page 23: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

THE META-INFORMATIVETHE META-INFORMATIVECENTERING THEORYCENTERING THEORY

(MIC)(MIC)

Page 24: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

What is Information in Grammar ?What is Information in Grammar ?What is Information in Grammar ?What is Information in Grammar ?

What linguists,What linguists, following the Prague School’s tradition, following the Prague School’s tradition, usually call usually call information,information, we named we named meta-informationmeta-information. .

In the age of unification of many social sciences under the label of cognitive sciences, it seems justied that the term information were used in the same way, at least, in linguistics and in information science.

Information is the semantic content of an utterance. Information is produced when properties an/or relations are established for entities.

Page 25: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Motivations of Old/New StatusMotivations of Old/New StatusMotivations of Old/New StatusMotivations of Old/New Status

Three kinds of motivations of Old and New meta-informative status:Three kinds of motivations of Old and New meta-informative status:

(a) (a) TThe communicative he communicative motivation is explicit and speech bound. The motivation is explicit and speech bound. The situation spoken about is either connected to another one mentioned situation spoken about is either connected to another one mentioned before (before (anaphoricanaphoric) or to be mentioned () or to be mentioned (cataphoriccataphoric) or it is a modal ) or it is a modal situation (ex. either reported or to be reported). situation (ex. either reported or to be reported).

(b) The (b) The cognitive cognitive motivation is related to the acquisition of knowledge. motivation is related to the acquisition of knowledge. Situations appear as (or are only presented as if they were) already Situations appear as (or are only presented as if they were) already known known (registered) or (registered) or unknown unknown (unregistered).(unregistered).

(c) The (c) The epistemic (epistemic (ontological ) ontological ) motivation depends on the knowledge motivation depends on the knowledge stored in long term memory; the situation spoken about is treated either as stored in long term memory; the situation spoken about is treated either as a class (a class (generic, general, habitual generic, general, habitual or or potentialpotential) or an instance () or an instance (specific, specific, particular, occasional particular, occasional or or actualactual). ).

Page 26: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Status MotivationStatus MotivationStatus MotivationStatus Motivation

Page 27: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

ATTENTION PROPERTIESATTENTION PROPERTIES

Selection

Orientation

http://www.icevi.org/publications/ICEVI-WC2002/papers/07-topic/07-ingsholt1.htm

Control

Page 28: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Centre of Attention in the MIC theoryCentre of Attention in the MIC theoryCentre of Attention in the MIC theoryCentre of Attention in the MIC theory

In linguistics, there was a need to define a general concept In linguistics, there was a need to define a general concept in order to capture what is common between the notions of in order to capture what is common between the notions of Subject, Object, Topic and Focus. In the MIC theory, this Subject, Object, Topic and Focus. In the MIC theory, this concept are called “Centre of Attention” (CA). It is considered concept are called “Centre of Attention” (CA). It is considered not only as a psychological phenomenon but also as not only as a psychological phenomenon but also as underlying segments of linguistic utterances.underlying segments of linguistic utterances.

No judgment may be stated without selecting at least one No judgment may be stated without selecting at least one Centre of Attention (CA).Centre of Attention (CA).

In the MIC theory, centering is defined as a structuring In the MIC theory, centering is defined as a structuring operation not only within a text (between utterances) but operation not only within a text (between utterances) but basically within the utterance limits.basically within the utterance limits.

Page 29: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Informative and Meta-Informative Informative and Meta-Informative Assignment FunctionsAssignment Functions

Informative and Meta-Informative Informative and Meta-Informative Assignment FunctionsAssignment Functions

inf(inf(rr)) = = ee μ-inf(μ-inf(cc)) = = rr

μ-infμ-inf

Role/Role/AnchorAnchor

PragmaticsPragmatics

ccSubject/ObjectSubject/Object

Centres of Centres of AttentionAttention

OntologyOntology

eeAgent/(Figure)Agent/(Figure)

LocationLocation

Participant/Participant/LocationLocation

infinf SemanticsSemantics

rrActive/PassiveActive/Passive

Page 30: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Semantic ContentSemantic Content

SIT frame: SIT frame: treattreat

(treating : (treating : “Mary”)) (treated :(treated : “Peter”))

SemanticSemanticLevelLevel

information

Utterance:Utterance: Mary treats Peter.

Page 31: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Syntactic Constituency as Meta-InformationSyntactic Constituency as Meta-InformationSyntactic Constituency as Meta-InformationSyntactic Constituency as Meta-Information

SIT : SIT : treat

(treating :(treating : “Mary” ))(treated :(treated : “Peter”))

Semantic LevelSemantic LevelInformation

Pragmatic levelPragmatic levelMeta-information

SubjectSubject PredicatePredicate ObjectObject

Page 32: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Predication and its ExtensionsPredication and its ExtensionsPredication and its ExtensionsPredication and its Extensions

SemanticsSemantics

information

meta-information PragmaticsPragmatics

SubjectGlobalPredicatePredicationPredication

Utterance : As for Mary, it is Peter whom she treats.

LocalPredicateLocalPredicate

ObjectObject

SIT : SIT : treat

(treating :(treating : “Mary” ))(treated :(treated : “Peter”))

TopicExtensionsExtensions

CommentBackground

Focus

meta-meta-information

Page 33: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Semantic and pragmatic levelsSemantic and pragmatic levels

Subject : (Predicate (Object))

Semantic levelSemantic levelInformation

Pragmatic levelPragmatic level

Meta-information

(median role : means)

““Mary treats Peter with aspirin.”

Utterance: Mary treats Peter with aspirinUtterance: Mary treats Peter with aspirin

The semantic role of the instrument only is expressed explicitly (“with”).The semantic role of the instrument only is expressed explicitly (“with”).

(active role) treats (passive role)

Page 34: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Structural SimilarityStructural Similaritybetween between BaseBase and Extended Utterences and Extended Utterences

Structural SimilarityStructural Similaritybetween between BaseBase and Extended Utterences and Extended Utterences

Subject

ObjectVerb

Predicate

Base utterance

Topic

Background Focus

Comment

Extended utterance

Global Aboutness

Local Aboutness

Global Aboutness

Local Aboutness

Global CA Global CA

Local CA Local CA

Page 35: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Meta-informativeMeta-informative pivots of discoursepivots of discourseMeta-informativeMeta-informative pivots of discoursepivots of discourse

Pragmatic UnitsPragmatic UnitsCentres of AttentionCentres of Attention

GlobalGlobal LocalLocal

BaseBase Utterance (Predication) Utterance (Predication) SubjectSubject ObjectObject

Extended Utterance (Extension)Extended Utterance (Extension) TopicTopic FocusFocus

Dialogue/Text (Discourse)Dialogue/Text (Discourse) GeneralGeneral

ThemeTheme

ParticularParticular

ThemeTheme

Page 36: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

BaseBase Utterances Utterances(orthogonal system)(orthogonal system)

SubjectSubject PredicatePredicate

SubjectSubject PredicatePredicate

“Old” Status

“New” Status

Page 37: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Extended UtterancesExtended Utterances(orthogonal system)(orthogonal system)

FocusFocus CommentComment

TopicTopic BackgroundBackground

“Old” Status

“New” Status

Page 38: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

BaseBase and Extended Utterances and Extended Utterances(orthogonal system)(orthogonal system)

SubjectSubject PredicatePredicate

SubjectSubject PredicatePredicateTopicTopic

Comment

Comment

Focus

Focus

Background

Background

“Old” Status

“New” Status

Page 39: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Implicit Subjects Implicit Subjects and and Topics Topics

FOCUSFOCUS

SUBJECTSUBJECT TOPICTOPIC

Explicature

Implicature

OBJECTOBJECT

Expression

Refinement

Page 40: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

CombinabilityCombinabilityof Centres of Attention with Semantic Rolesof Centres of Attention with Semantic Roles

CombinabilityCombinabilityof Centres of Attention with Semantic Rolesof Centres of Attention with Semantic Roles

TopicTopic Subject Subject Active Active rolerole

FocusFocus Object Object Passive Passive rolerole

Page 41: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Meta-informative paraphrasesMeta-informative paraphrasesMeta-informative paraphrasesMeta-informative paraphrases1a. MaryMary treats Peter. (Active voice + [Subject || Active r.] + [Object || Passive r.])1b. Peter is treated by Mary.(Passive voice + [Subject || Passive r.] + [Object || Active r.] )2a. As for Mary, she treats Peter.(Active voice + [Topic || Subject || Active r.] + [Object || Passive r.])2b. As for Peter, he is treated by Mary.(Passive voice + [Topic || Subject || Passive r.] + [Object || Active r.] )3a. As for Mary ,it is Peter whom she treats . (Active voice + [Topic || Subject || Active r.] + [Focus || Object || Passive r.])3b. As for Peter, it is Mary who treats him.(Active voice + [Topic || Object || Passive r.] + [Focus || Subject || Active r.])4a. As for Peter, it is by Mary that he is treated . (Passive voice + [Topic || Subject || Passive r.] + [Focus || Object || Active r.])4b. ?? As for Mary, it is by her that Peter is treated . (Passive voice + [Topic || Object || Active r.] + [Focus || Subject || Passive r.])etc.

Page 42: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Homogeneous and HeterogeneousHomogeneous and Heterogeneousmeta-informative statusmeta-informative status

Homogeneous and HeterogeneousHomogeneous and Heterogeneousmeta-informative statusmeta-informative status

BaseBase Utterance (Schemas) Utterance (Schemas) BaseBase Utterance (Examples) Utterance (Examples)

(New) Subject : (New) Predicate(New) Subject : (New) Predicate #1 A new satellite has been launched #1 A new satellite has been launched today.today.

(Old) Subject : (Old) Predicate(Old) Subject : (Old) Predicate #2 Satellites turn around the Earth.#2 Satellites turn around the Earth.

Extended Utterance (Schemas)Extended Utterance (Schemas) Extended Utterance (Examples)Extended Utterance (Examples)

(Old) Topic : (New) Comment(Old) Topic : (New) Comment #3 As for the satellite X03, it has been #3 As for the satellite X03, it has been destroyed by a meteorite.destroyed by a meteorite.

(New) Focus : (Old) Background(New) Focus : (Old) Background #4 It is #4 It is the the satellite X03 which was satellite X03 which was destroyed today. destroyed today.

Page 43: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Japanese base utterances(with ‘wa’ and ‘ga’ particles)

Japanese base utterances(with ‘wa’ and ‘ga’ particles)

SUBJECT (Old status) + Predicate (Old status)SUBJECT (Old status) + Predicate (Old status)地球は太陽の周囲を回転する。地球は太陽の周囲を回転する。 The Earth goes around the Sun.The Earth goes around the Sun.Chikyuu wa taiyou no shui wo kaiten-suru.Chikyuu wa taiyou no shui wo kaiten-suru.Earth WA(Nom) Nom Sun NO(Gen) periphery WO(Acc) turn-aroundEarth WA(Nom) Nom Sun NO(Gen) periphery WO(Acc) turn-around

SUBJECT (New status) + Predicate (New status)SUBJECT (New status) + Predicate (New status)夜の底が白くなった。夜の底が白くなった。 (( 川端康成川端康成 ) The profound night became ) The profound night became white.white.Yoru no soko ga shiroku natta. (Kawabata Yasunari).Yoru no soko ga shiroku natta. (Kawabata Yasunari).Night NO(Gen) bottom GA(Nom) white became Past.Night NO(Gen) bottom GA(Nom) white became Past.

SUBJECT (Old status) + Predicate (Old status)SUBJECT (Old status) + Predicate (Old status)地球は太陽の周囲を回転する。地球は太陽の周囲を回転する。 The Earth goes around the Sun.The Earth goes around the Sun.Chikyuu wa taiyou no shui wo kaiten-suru.Chikyuu wa taiyou no shui wo kaiten-suru.Earth WA(Nom) Nom Sun NO(Gen) periphery WO(Acc) turn-aroundEarth WA(Nom) Nom Sun NO(Gen) periphery WO(Acc) turn-around

SUBJECT (New status) + Predicate (New status)SUBJECT (New status) + Predicate (New status)夜の底が白くなった。夜の底が白くなった。 (( 川端康成川端康成 ) The profound night became ) The profound night became white.white.Yoru no soko ga shiroku natta. (Kawabata Yasunari).Yoru no soko ga shiroku natta. (Kawabata Yasunari).Night NO(Gen) bottom GA(Nom) white became Past.Night NO(Gen) bottom GA(Nom) white became Past.

Page 44: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Centering and ConstituencyCentering and ConstituencyCentering and ConstituencyCentering and Constituency

As Centering is governed by constituency, the Canonic As Centering is governed by constituency, the Canonic Word Order - like in Context Free Grammars - quite naturally Word Order - like in Context Free Grammars - quite naturally depends on it.depends on it.

It is especially crucial as regards the meta-informative It is especially crucial as regards the meta-informative level because the word order of extended utterances (higher level because the word order of extended utterances (higher levels) follows additional rules which depend on the levels) follows additional rules which depend on the hierarachical relationship of combined centers of attention hierarachical relationship of combined centers of attention (CA).(CA).

- - - - - - - - - - - - - - - - - - -- - - - - - - - - - - - - - - - - - -Note however, we de not pretend that the constituency Note however, we de not pretend that the constituency

principle underlies only the meta-informative level of principle underlies only the meta-informative level of language use.language use.

Page 45: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Canonic Word Order and CompositionalityCanonic Word Order and CompositionalityCanonic Word Order and CompositionalityCanonic Word Order and Compositionality

John loves Mary.John loves Mary. Isabelle Tellier showed that syntactic structures can be considered

as a direct consequence of a need to functionnally combine meanings.Let us define: love2’ as a binary (arity 2) predicate, John’ and

Mary’ as two logical individual constants.Let us also define two oriented functional applications h1 and h2:• for any couple of semantic expressions a and b, h1(a, b) = a(b) = a/b• for any couple of semantic expressions a and b, h2(a, b) = b(a) = b\a

Thus, the logical relationship loves2’(Mary’) (John’)loves2’(Mary’) (John’) may be mapped onto each of the following constituency (phrase) structures:

Isabelle Tellier, « Semantic-Driven Emergence of Syntax : The Principle of Compositionality upside-down »http://www.univ-orleans.fr/lifo/Members/Isabelle.Tellier/recherche.html

h1(h2(Mary', love2'), John')) : OVSh1(h2(Mary', love2'), John')) : OVSh1(h1(love2', Mary'), John') : VOSh1(h1(love2', Mary'), John') : VOS

h2(John', h1(love2', Mary')) : SVOh2(John', h1(love2', Mary')) : SVOh2(John', h2(Mary', love2')) : SOVh2(John', h2(Mary', love2')) : SOV

Page 46: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Canonic Word Order and CompositionalityCanonic Word Order and Compositionality(two special cases)(two special cases)

Canonic Word Order and CompositionalityCanonic Word Order and Compositionality(two special cases)(two special cases)

John loves Mary. John loves Mary. loves(Mary) (John).loves(Mary) (John).

The OSV and VSO type sentences need special treatment explicitly The OSV and VSO type sentences need special treatment explicitly stating word orders :stating word orders :

•love2'(Mary')(John')= h2(Mary', h2(John', λxλy.love2'(y)(x)): OSVlove2'(Mary')(John')= h2(Mary', h2(John', λxλy.love2'(y)(x)): OSV•love2'(Mary')(John')= h1(h1(λxλy.love2'(y)(x,)John'), Mary') : VSOlove2'(Mary')(John')= h1(h1(λxλy.love2'(y)(x,)John'), Mary') : VSO

Isabelle Tellier, « Semantic-Driven Emergence of Syntax : The Principle of Compositionality upside-down »http://www.univ-orleans.fr/lifo/Members/Isabelle.Tellier/recherche.html

Page 47: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Centering and Canonic Word OrderCentering and Canonic Word OrderCentering and Canonic Word OrderCentering and Canonic Word Order

The following two orders OSV and VSO contradict the The following two orders OSV and VSO contradict the Principle of Constituency. This is due to the fact that the Principle of Constituency. This is due to the fact that the constituents which correspond to the Global Centres of constituents which correspond to the Global Centres of Attention cannot sit in the middle place in the Canonic Word Attention cannot sit in the middle place in the Canonic Word OrderOrder. .

**OSVOSV

OO SS VV VV

**VSOVSO

SS OO

Thus, the Meta-Informative Centering theory (MIC), now Thus, the Meta-Informative Centering theory (MIC), now part of the Distributed Grammar, has a predictive potential.part of the Distributed Grammar, has a predictive potential.

Page 48: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Centering and Word OrderCentering and Word Order(sample 1)(sample 1)

Centering and Word OrderCentering and Word Order(sample 1)(sample 1)

Word Order Frequencies according to “CHILDES” “CHILDES” DatabaseDatabase

HOFFMAN Beryl (1996) "Word Order, Information Structure and Centering in Turkish", in 'Centering in Discourse', eds. Ellen Prince, Aravind Joshi and Marilyn Walker, Oxford

http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.52.8937

SOV 48 %OSV 8 %SVO 25 %OVS 13 %VSO 6 %VOS < 1 %

Page 49: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

Centering and Word OrderCentering and Word Order(sample 2)(sample 2)

Centering and Word OrderCentering and Word Order(sample 2)(sample 2)

Subject-object-verb (SOV) – 497 (40,47 %) Subject-object-verb (SOV) – 497 (40,47 %) Subject-verb-object (SVO) – 436 (35,50 %)Subject-verb-object (SVO) – 436 (35,50 %)Verb-subject-object (VSO) – 85 (6,92 %)Verb-subject-object (VSO) – 85 (6,92 %)Verb-object-subject (VOS) – 26 (2,12 %)Verb-object-subject (VOS) – 26 (2,12 %)Object-verb-subject (OVS) – 9 (0,73 %)Object-verb-subject (OVS) – 9 (0,73 %)Object-subject-verb (OSV) – 4 (0,36 %)Object-subject-verb (OSV) – 4 (0,36 %)Lacking dominant word order – 171 (13,93 %)Lacking dominant word order – 171 (13,93 %) total: 1228total: 1228 languages languages

http://wals.info/feature/description/81

Word Order Frequenciesaccording to “The World Atlas of Language Structures

Online”

Page 50: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

CONCLUSIONCONCLUSIONIn order to communicate the human brain processes (produces/understands) linguistic units (utterances). Although there is a number of different linear forms of utterances (which depend, among others, on the valency schemata), I claim that it is necessary to model the utterance meaning (definable jointly in the light of semantics, pragmatics and praxematics) as an open-ended aggregate of conceptual representations.

These representations are the result of activations of various informational devices within the multi-dimensional space of memorised knowledge. The relative complicatedness of the Distributed Grammar model reflects the complexity of the environment in which the man needs to interact in an intelligent way with other men in order to survive.

Page 51: Towards Distributed Grammar motivations and issues of representation André WLODARCZYK

© André WLODARCZYK

http://www.celta.paris-sorbonne.fr