Dialogue Modelling, Language Processing Dynamics and Linguistic - - PowerPoint PPT Presentation

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Dialogue Modelling, Language Processing Dynamics and Linguistic Knowledge

Eleni Gregoromichelaki, Ruth Kempson, Gregory Mills, Patrick Healey The Dynamics of Conversational Dialogue (DynDial) ESRC-RES-062-23-0962 www.kcl.ac.uk/research/groups/ds

Gregoromichelaki et al Osnabr¨ uck 19/09/09 1/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction Dynamic Syntax: incremental structure/content growth Grammar design

Gregoromichelaki et al Osnabr¨ uck 19/09/09 2/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Outline

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction Dynamic Syntax: incremental structure/content growth Grammar design

Gregoromichelaki et al Osnabr¨ uck 19/09/09 3/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Dialogue data and linguistic knowledge

◮ linguistic knowledge provides constraints on linguistic

processing (Bosch 2008)

• linguistic knowledge: (often) characterised in abstract static

terms

• linguistic processing: immediacy, incrementality, crossmodality

(see Marslen-Wilson & Tyler 1980, Altmann & Steedman 1988)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 4/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Dialogue data and linguistic knowledge

◮ linguistic knowledge provides constraints on linguistic

processing (Bosch 2008)

• linguistic knowledge: (often) characterised in abstract static

terms

• linguistic processing: immediacy, incrementality, crossmodality

(see Marslen-Wilson & Tyler 1980, Altmann & Steedman 1988)

◮ ellipsis resolution requires essentially all the above

• a unified story requires processor properties included in the

theory of linguistic knowledge (Kempson et al 2009, cf Bosch 2008)

• dialogue data: separating linguistic knowledge from processing

results in view of dialogue as “degenerate” language

Gregoromichelaki et al Osnabr¨ uck 19/09/09 4/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Dialogue data and linguistic knowledge

◮ linguistic knowledge provides constraints on linguistic

processing (Bosch 2008)

• linguistic knowledge: (often) characterised in abstract static

terms

• linguistic processing: immediacy, incrementality, crossmodality

(see Marslen-Wilson & Tyler 1980, Altmann & Steedman 1988)

◮ ellipsis resolution requires essentially all the above

• a unified story requires processor properties included in the

theory of linguistic knowledge (Kempson et al 2009, cf Bosch 2008)

• dialogue data: separating linguistic knowledge from processing

results in view of dialogue as “degenerate” language

◮ language-as-action and language-as-product: shifting

boundaries of grammar and pragmatics

Gregoromichelaki et al Osnabr¨ uck 19/09/09 4/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

What dialogue data tell us

◮ The evolving, structural nature of NL context dependence

• elliptical phenomena reveal the nature of context, in

monologue and dialogue equally

• context-dependent phenomena as invariably straddling

syntax/semantics/pragmatics (see e.g. Cooper and Ginzburg

2004)

• context as evolving record of structured representations

(Hamm et al 2006, etc) but richer than standard DRT

Gregoromichelaki et al Osnabr¨ uck 19/09/09 5/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

What dialogue data tell us

◮ The evolving, structural nature of NL context dependence

• elliptical phenomena reveal the nature of context, in

monologue and dialogue equally

• context-dependent phenomena as invariably straddling

syntax/semantics/pragmatics (see e.g. Cooper and Ginzburg

2004)

• context as evolving record of structured representations

(Hamm et al 2006, etc) but richer than standard DRT

◮ In conversation, inference based on mutual

knowledge/common ground is not a prerequisite for communication

• the content of “intentions” may emerge as a result of

communication, instead of guiding it

Gregoromichelaki et al Osnabr¨ uck 19/09/09 5/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

What dialogue data tell us

◮ The evolving, structural nature of NL context dependence

• elliptical phenomena reveal the nature of context, in

monologue and dialogue equally

• context-dependent phenomena as invariably straddling

syntax/semantics/pragmatics (see e.g. Cooper and Ginzburg

2004)

• context as evolving record of structured representations

(Hamm et al 2006, etc) but richer than standard DRT

◮ In conversation, inference based on mutual

knowledge/common ground is not a prerequisite for communication

• the content of “intentions” may emerge as a result of

communication, instead of guiding it

◮ Linguistic knowledge: the update dynamics of communication

• word by word incrementality within a grammar system
• NL grammars as mechanisms for communicative interaction

Gregoromichelaki et al Osnabr¨ uck 19/09/09 5/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Outline

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction Dynamic Syntax: incremental structure/content growth Grammar design

Gregoromichelaki et al Osnabr¨ uck 19/09/09 6/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

The new challenge of conversational dialogue Clark 1996,

Pickering & Garrod 2004, Ginzburg 2009

◮ Dialogue contains a high proportion of ellipsis hence

radically context-dependent processing

Gregoromichelaki et al Osnabr¨ uck 19/09/09 7/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

The new challenge of conversational dialogue Clark 1996,

Pickering & Garrod 2004, Ginzburg 2009

◮ Dialogue contains a high proportion of ellipsis hence

radically context-dependent processing but

Gregoromichelaki et al Osnabr¨ uck 19/09/09 7/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

The new challenge of conversational dialogue Clark 1996,

Pickering & Garrod 2004, Ginzburg 2009

◮ Dialogue contains a high proportion of ellipsis hence

radically context-dependent processing but

◮ Dialogue primary site of language use ◮ Conversational dialogue sole basis for data from which

children learn language

Gregoromichelaki et al Osnabr¨ uck 19/09/09 7/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

The new challenge of conversational dialogue Clark 1996,

Pickering & Garrod 2004, Ginzburg 2009

◮ Dialogue contains a high proportion of ellipsis hence

radically context-dependent processing but

◮ Dialogue primary site of language use ◮ Conversational dialogue sole basis for data from which

children learn language so

Gregoromichelaki et al Osnabr¨ uck 19/09/09 7/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

The new challenge of conversational dialogue Clark 1996,

Pickering & Garrod 2004, Ginzburg 2009

◮ Dialogue contains a high proportion of ellipsis hence

radically context-dependent processing but

◮ Dialogue primary site of language use ◮ Conversational dialogue sole basis for data from which

children learn language so

◮ Dialogue as merely degenerate language use or a unified

analysis of ellipsis (in monologue/dialogue) as context-dependence?

Gregoromichelaki et al Osnabr¨ uck 19/09/09 7/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Split-utterances

◮ Collaboratively constructed discourse

(1) Context: Friends of the Earth club meeting A: So what is that? Is that er... booklet or something? B: It’s a book C: Book B: Just ... talking about al you know alternative D: On erm... renewable yeah B: energy really I think...... A: Yeah [BNC:D97]

Gregoromichelaki et al Osnabr¨ uck 19/09/09 8/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Split-utterances

◮ Collaboratively constructed discourse

(1) Context: Friends of the Earth club meeting A: So what is that? Is that er... booklet or something? B: It’s a book C: Book B: Just ... talking about al you know alternative D: On erm... renewable yeah B: energy really I think...... A: Yeah [BNC:D97]

◮ Where do utterance boundaries occur?

(2) A: We’re going to Marlborough B: to see Granny.

Gregoromichelaki et al Osnabr¨ uck 19/09/09 8/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Developing utterances (together) in dialogue

◮ Speaker/hearer exchange of roles across all syntactic

dependencies (Purver et al 2009):

(3) A: Have you read ... B: any of your chapters? Not yet. (4) Gardener: I shall need the mattock. Home-owner: The... Gardener: mattock. For breaking up clods of earth.[BNC]

Gregoromichelaki et al Osnabr¨ uck 19/09/09 9/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Developing utterances (together) in dialogue

◮ Speaker/hearer exchange of roles across all syntactic

dependencies (Purver et al 2009):

(3) A: Have you read ... B: any of your chapters? Not yet. (4) Gardener: I shall need the mattock. Home-owner: The... Gardener: mattock. For breaking up clods of earth.[BNC]

◮ A language game emerging at earliest stages of language

acquisition

(5) Carer: Old McDonald had a farm... On that farm he had a Child: cow.

Gregoromichelaki et al Osnabr¨ uck 19/09/09 9/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Guessing intentions?

◮ Interruptions possible before proposition-intention fixable:

(6) A. They X-rayed me, and took a urine sample, took a blood sample. Er, the doctor B: Chorlton? A: Chorlton, mhm, he examined me....... [BNC]

Gregoromichelaki et al Osnabr¨ uck 19/09/09 10/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Guessing intentions?

◮ Interruptions possible before proposition-intention fixable:

(6) A. They X-rayed me, and took a urine sample, took a blood sample. Er, the doctor B: Chorlton? A: Chorlton, mhm, he examined me....... [BNC]

◮ Intentions emerge/develop during dialogue:

(7) Daughter: Oh here dad, a good way to get those corners out Dad: is to stick yer finger inside. Daughter: well, that’s one way (Lerner 1991)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 10/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Guessing intentions?

◮ Interruptions possible before proposition-intention fixable:

(6) A. They X-rayed me, and took a urine sample, took a blood sample. Er, the doctor B: Chorlton? A: Chorlton, mhm, he examined me....... [BNC]

◮ Intentions emerge/develop during dialogue:

(7) Daughter: Oh here dad, a good way to get those corners out Dad: is to stick yer finger inside. Daughter: well, that’s one way (Lerner 1991)

◮ Utterances may be multi-functional:

(8) A: Are you left or B: Right-handed (9) M: It’s generated with a handle and J: Wound round? [BNC] M: Yes, wind them round and this should, should generate a charge

Gregoromichelaki et al Osnabr¨ uck 19/09/09 10/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Dialogue data: challenges

◮ Dialogue should be difficult:

Gregoromichelaki et al Osnabr¨ uck 19/09/09 11/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Dialogue data: challenges

◮ Dialogue should be difficult:

• highly elliptical

Gregoromichelaki et al Osnabr¨ uck 19/09/09 11/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Dialogue data: challenges

◮ Dialogue should be difficult:

• highly elliptical
• multifunctionality/underspecification poses challenge of

defining syntactic/semantic types for a range of fragment types/speech acts (with incomplete information in the signal): clarifications, acknowledgements, answers, completions...

Gregoromichelaki et al Osnabr¨ uck 19/09/09 11/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Dialogue data: challenges

◮ Dialogue should be difficult:

• highly elliptical
• multifunctionality/underspecification poses challenge of

defining syntactic/semantic types for a range of fragment types/speech acts (with incomplete information in the signal): clarifications, acknowledgements, answers, completions...

• radical incrementality poses challenge for all sentence-based

grammar formalisms

Gregoromichelaki et al Osnabr¨ uck 19/09/09 11/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Dialogue data: challenges

◮ Dialogue should be difficult:

• highly elliptical
• multifunctionality/underspecification poses challenge of

defining syntactic/semantic types for a range of fragment types/speech acts (with incomplete information in the signal): clarifications, acknowledgements, answers, completions...

• radical incrementality poses challenge for all sentence-based

grammar formalisms

• poses challenge of modelling the interlocutor’s mind (with

incomplete information)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 11/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Dialogue data: challenges

◮ Dialogue should be difficult:

• highly elliptical
• multifunctionality/underspecification poses challenge of

defining syntactic/semantic types for a range of fragment types/speech acts (with incomplete information in the signal): clarifications, acknowledgements, answers, completions...

• radical incrementality poses challenge for all sentence-based

grammar formalisms

• poses challenge of modelling the interlocutor’s mind (with

incomplete information)

• switching parsing/producing tasks before propositional

content is available:

• highly incremental understanding
• highly incremental planning what to say (while listening)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 11/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Dialogue data: challenges

◮ Dialogue should be difficult:

• highly elliptical
• multifunctionality/underspecification poses challenge of

defining syntactic/semantic types for a range of fragment types/speech acts (with incomplete information in the signal): clarifications, acknowledgements, answers, completions...

• radical incrementality poses challenge for all sentence-based

grammar formalisms

• poses challenge of modelling the interlocutor’s mind (with

incomplete information)

• switching parsing/producing tasks before propositional

content is available:

• highly incremental understanding
• highly incremental planning what to say (while listening)

◮ How can children do it so easily?

these are the sole data on which children acquire language

Gregoromichelaki et al Osnabr¨ uck 19/09/09 11/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Dialogue and standard syntax/semantics models

◮ Denotational semantics (NB: not of the Bosch kind) is

externalist, unrelated to all cognitive considerations, and, without structure, inadequate for explanation of dialogue ellipsis.

• (naive) GQ semantics not predicting psycholinguistic

results/dialogue phenomena (Bosch 2008, Purver & Ginzburg 2004)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 12/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Dialogue and standard syntax/semantics models

◮ Denotational semantics (NB: not of the Bosch kind) is

externalist, unrelated to all cognitive considerations, and, without structure, inadequate for explanation of dialogue ellipsis.

• (naive) GQ semantics not predicting psycholinguistic

results/dialogue phenomena (Bosch 2008, Purver & Ginzburg 2004)

◮ Non-incremental, context-insensitive grammars:

◮ inadequacy of head-driven models: fragments can be resolved

before emergence of head-projected structure.

◮ speaker/listener switching deeply problematic for all

sentence-based models.

Gregoromichelaki et al Osnabr¨ uck 19/09/09 12/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Dialogue and standard syntax/semantics models

◮ Denotational semantics (NB: not of the Bosch kind) is

externalist, unrelated to all cognitive considerations, and, without structure, inadequate for explanation of dialogue ellipsis.

• (naive) GQ semantics not predicting psycholinguistic

results/dialogue phenomena (Bosch 2008, Purver & Ginzburg 2004)

◮ Non-incremental, context-insensitive grammars:

◮ inadequacy of head-driven models: fragments can be resolved

before emergence of head-projected structure.

◮ speaker/listener switching deeply problematic for all

sentence-based models.

◮ Concept of context needed for ellipsis neither denotational

nor static: context involves incremental structural update.

◮ Arbitrary sentence parts can be context for subsequent

(elliptical) fragment (Purver et al 2009)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 12/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Ellipsis and current syntax/semantics models

◮ Elliptical forms need syntactic characterisation but involve

postulating multiple phenomena/ambiguities: stripping, gapping, sluicing, antecedent-contained ellipsis. No basis for parallelism (Fiengo & May 1994, etc.)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 13/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Ellipsis and current syntax/semantics models

◮ Elliptical forms need syntactic characterisation but involve

postulating multiple phenomena/ambiguities: stripping, gapping, sluicing, antecedent-contained ellipsis. No basis for parallelism (Fiengo & May 1994, etc.)

◮ Semantic accounts explain parallelism but not fully

successfully despite over-powerful mechanism (Dalrymple et al

1991) with no basis for morpho-syntactic or syntactic

constraints (see Morgan 1973, Webber 1979, Steedman 2000)

(1) Hat Kim nicht den Brief geschrieben? Nein Ich/*Mich Did Kim not write the letter ? No, INOM (2) *John interviewed everyone who Bill knew the woman who had.

Gregoromichelaki et al Osnabr¨ uck 19/09/09 13/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Ellipsis and current syntax/semantics models

◮ Elliptical forms need syntactic characterisation but involve

postulating multiple phenomena/ambiguities: stripping, gapping, sluicing, antecedent-contained ellipsis. No basis for parallelism (Fiengo & May 1994, etc.)

◮ Semantic accounts explain parallelism but not fully

successfully despite over-powerful mechanism (Dalrymple et al

1991) with no basis for morpho-syntactic or syntactic

constraints (see Morgan 1973, Webber 1979, Steedman 2000)

(1) Hat Kim nicht den Brief geschrieben? Nein Ich/*Mich Did Kim not write the letter ? No, INOM (2) *John interviewed everyone who Bill knew the woman who had.

◮ Pragmatic accounts are partial and presume an independent

grammar (Carston 2002, Stainton 2006)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 13/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Dialogue/ellipsis and pragmatic models

◮ Standard assumption: understanding involves recognition of speaker’s

intentions, grounded in mutual knowledge/common ground (Grice, Sperber & Wilson, Clark etc etc)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 14/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Dialogue/ellipsis and pragmatic models

◮ Standard assumption: understanding involves recognition of speaker’s

intentions, grounded in mutual knowledge/common ground (Grice, Sperber & Wilson, Clark etc etc) but

◮ Mutual knowledge paradox/common ground intractability

Gregoromichelaki et al Osnabr¨ uck 19/09/09 14/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Dialogue/ellipsis and pragmatic models

◮ Standard assumption: understanding involves recognition of speaker’s

intentions, grounded in mutual knowledge/common ground (Grice, Sperber & Wilson, Clark etc etc) but

◮ Mutual knowledge paradox/common ground intractability ◮ Speech act recognition in dialogue shown to be derivative (Conversation

Analysis, Levinson 1983, Pickering & Garrod 2004)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 14/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Dialogue/ellipsis and pragmatic models

◮ Standard assumption: understanding involves recognition of speaker’s

intentions, grounded in mutual knowledge/common ground (Grice, Sperber & Wilson, Clark etc etc) but

◮ Mutual knowledge paradox/common ground intractability ◮ Speech act recognition in dialogue shown to be derivative (Conversation

Analysis, Levinson 1983, Pickering & Garrod 2004)

◮ Radical incrementality of parsing/generation in dialogue: recognition of

propositional intention appears not to be basic

Gregoromichelaki et al Osnabr¨ uck 19/09/09 14/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Dialogue/ellipsis and pragmatic models

◮ Standard assumption: understanding involves recognition of speaker’s

intentions, grounded in mutual knowledge/common ground (Grice, Sperber & Wilson, Clark etc etc) but

◮ Mutual knowledge paradox/common ground intractability ◮ Speech act recognition in dialogue shown to be derivative (Conversation

Analysis, Levinson 1983, Pickering & Garrod 2004)

◮ Radical incrementality of parsing/generation in dialogue: recognition of

propositional intention appears not to be basic so

◮ What grounds utterance understanding?

Gregoromichelaki et al Osnabr¨ uck 19/09/09 14/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Dialogue/ellipsis and pragmatic models

◮ Standard assumption: understanding involves recognition of speaker’s

intentions, grounded in mutual knowledge/common ground (Grice, Sperber & Wilson, Clark etc etc) but

◮ Mutual knowledge paradox/common ground intractability ◮ Speech act recognition in dialogue shown to be derivative (Conversation

Analysis, Levinson 1983, Pickering & Garrod 2004)

◮ Radical incrementality of parsing/generation in dialogue: recognition of

propositional intention appears not to be basic so

◮ What grounds utterance understanding? ◮ Children can engage in sophisticated dialogue exchanges before

mind-reading capacity emerges:

• do children communicate in the same way as adults? (Breheny 2006)
• since children acquire systematic sentence-building abilities from

conversational dialogue, how do they do it?

Gregoromichelaki et al Osnabr¨ uck 19/09/09 14/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Dialogue/ellipsis and psycholinguistic models

◮ Coordination in dialogue emergent without necessary

calculation of common ground/speaker’s intention (Pickering &

Garrod 2004, Keysar, Barr et al, Horton & Gerrig)

◮ perspective-adjustment/audience design experiments (Keysar,

Barr, Horton, Gerrig et al): egocentrism

• no default explicit metarepresentation of interlocutor’s mental

state “common ground is a functionally distinct process that belongs to an ’adjustment’ stage of processing, but that it imposes no constraint on production or comprehension per se” (Barr & Keysar: 904)

• a mechanistic model of apparent common ground

computation based on more basic memory mechanisms

◮ maze game experiments (Mills 2007; Mills & Gregoromichelaki

in prep): rich structure of exchange guides interpretation and coordination

Gregoromichelaki et al Osnabr¨ uck 19/09/09 15/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Chat-tool design

http://imc.dcs.qmul.ac.uk/diet/index.php/Main Page

Gregoromichelaki et al Osnabr¨ uck 19/09/09 16/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Maze game through chat-tool

Gregoromichelaki et al Osnabr¨ uck 19/09/09 17/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Fragments in maze game

◮ At late stages of interaction sequences contract radically and

become highly elliptical (telescoping): (1) A: 4,5 2,6 1,4 B: 1,2 3,4 7,1 A: 1,2 B: 4,5 A: 1,2 from Mills (2007)

◮ Contraction not only of referring expressions, (e.g. Krauss and

Weinheimer 1975): (2) A: 1, 2 “1 across from the left and 2 down”

◮ But also of sequences of instructions (sequential ordering

provides interpretation of what turn “is doing”)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 18/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Fragments in maze game

A: 4,5 2,6 1,4 B: 1,2 3,4 7,1 A: 1,2 B: 4,5 A: 1,2

Gregoromichelaki et al Osnabr¨ uck 19/09/09 19/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Unpacking the fragments (1)

A: 1,2 2,6 1,4 (I have 3 switches: Counting in the same way that we were counting, starting at 0 as the edge of the maze, and counting down from the top of the maze, also from 0, 1 square along and 2 squares down, 2 squares along and 6 down, 1 along and 4 down. If you go to any of them, they will open all of my gates, and I’ll go through and then you can guide me to yours to do the same. Can you get to any of these switches?) B: 4,5 3,4 7,1 (as above)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 20/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Unpacking the fragments (2)

A: 1,2 (Out of those three switches, 1 along and 2 across, counting from left to [......] I can get to this one. I’m going there right now, and can you please tell me whether your gates are open?) B: 4,5 (I can get to 4,5 and I am on it) A: 1,2 (So now I am blocked by a gate between me and my goal, I’m stuck at 1,2 [...] so could you please tell me another gate that is close to this location for me to go to,

• r if you can get to one
• f my switches directly,

then please go there to open my gates, and then I will go to my goal.)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 21/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Chat-tool data: interpetation in dialogue

◮ interpretation of fragments in chat-tool data diagnostic of

underspecification and plasticity of language to make it adaptable to novel situations:

◮ Interaction in context allows association of words with ad hoc

concepts to suit the situation (e.g. Carston 1998; Cooper & Ranta 2006; Larson 2008; Bosch this workshop)

◮ Semantic ontologies/interpretations arise during interaction

rather than a priori (Healey & Mills 2006; Mills & Healey 2008)

◮ Fragments interpretable according to (routinised) structure

imposed by the task/participants

Gregoromichelaki et al Osnabr¨ uck 19/09/09 22/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

The emergence of intentions

◮ Explicit negotiation more likely to impede at initial rounds of

tasks (Mills 2007; Healey 2007)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 23/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

The emergence of intentions

◮ Explicit negotiation more likely to impede at initial rounds of

tasks (Mills 2007; Healey 2007)

(1) Negotiation of descriptive schema: A: OK, let’s use rows and columns, mine is row 3 and column 6 (explicit attempt to introduce matrix scheme) B: Mine is on the sticking out bit A: Mine is underneath that (2) Negotiation of sequence of actions to solve the maze: A: OK, first you’ve got to tell me where to go and then I can go through B: where are your switches? A: tell me where to go and then I can get through B: I’m blocked by the gate in front of me

Gregoromichelaki et al Osnabr¨ uck 19/09/09 23/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

The emergence of intentions

◮ clarifications interpreted as concerning intentions at late

rounds of maze-game

Gregoromichelaki et al Osnabr¨ uck 19/09/09 24/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

The emergence of intentions

◮ clarifications interpreted as concerning intentions at late

rounds of maze-game as task experience increases, participants’ responses to clarification requests are disambiguated significantly more frequently as concerning “intentions” (Mills 2007; Mills & Gregoromichelaki in prep)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 24/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

The emergence of intentions

◮ clarifications interpreted as concerning intentions at late

rounds of maze-game as task experience increases, participants’ responses to clarification requests are disambiguated significantly more frequently as concerning “intentions” (Mills 2007; Mills & Gregoromichelaki in prep)

First few mazes: (1) A: Go to the top right of the maze Server: why? A: dunno/no idea (2) A: Can you get to the top of the maze? B: Why? A: Can you get to the top of the maze / Try it

Gregoromichelaki et al Osnabr¨ uck 19/09/09 24/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

The emergence of intentions

(3) Later on: A: Go to the 5th row, 3rd square Server: why? A: That’s where my switch is (4) A: 5, 6 Server: what?/5? A: because you’ve got to go there/you asked me to go there

(see e.g. Drew 1997; cf Purver 2004; Ginzburg & Cooper 2004)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 25/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

The emergence of intentions

◮ there are no clear underlying intentions/common ground

computation at work (even in task-specific dialogue)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 26/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

The emergence of intentions

◮ there are no clear underlying intentions/common ground

computation at work (even in task-specific dialogue)

◮ rather the emergence of (’routinised’) structure guides

efficient coordination

Gregoromichelaki et al Osnabr¨ uck 19/09/09 26/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

The emergence of intentions

◮ there are no clear underlying intentions/common ground

computation at work (even in task-specific dialogue)

◮ rather the emergence of (’routinised’) structure guides

efficient coordination

◮ Conversational Analysis: the function utterances perform are

due in large part to the place they occupy within specific conversational sequences

Gregoromichelaki et al Osnabr¨ uck 19/09/09 26/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

The emergence of intentions

◮ there are no clear underlying intentions/common ground

computation at work (even in task-specific dialogue)

◮ rather the emergence of (’routinised’) structure guides

efficient coordination

◮ Conversational Analysis: the function utterances perform are

due in large part to the place they occupy within specific conversational sequences

◮ “joint-intentions” emergent from conversation (cf Clark 1996)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 26/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

◮ If we revise the concept of utterance understanding,

dropping the necessary recovery of the proposition “which the speaker could have intended”

• n the basis of some pre-established

“mutual knowledge”/“common ground”

◮ what does utterance interpretation amount to?

Gregoromichelaki et al Osnabr¨ uck 19/09/09 27/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Language as action model: interactional approach

* “speakers recraft their utterances mid-stream, taking into account the responses, or more often the lack of them, from recipients . . . As a result, what is produced is actually a joint production, which can hardly correspond to the speaker’s own initial intention or goal.” (Goodwin 1979; 1981)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 28/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Language as action model: interactional approach

* “speakers recraft their utterances mid-stream, taking into account the responses, or more often the lack of them, from recipients . . . As a result, what is produced is actually a joint production, which can hardly correspond to the speaker’s own initial intention or goal.” (Goodwin 1979; 1981) * communication accomplished via a trial-and-error process (Arundale 2008)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 28/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Language as action model: interactional approach

* “speakers recraft their utterances mid-stream, taking into account the responses, or more often the lack of them, from recipients . . . As a result, what is produced is actually a joint production, which can hardly correspond to the speaker’s own initial intention or goal.” (Goodwin 1979; 1981) * communication accomplished via a trial-and-error process (Arundale 2008)

◮ Ginzburg et al: Dialogue understanding involves

modelling distinct fragment-categories via assigning distinct syntactic/semantic types (Ginzburg 2009)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 28/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction

Language as action model: interactional approach

* “speakers recraft their utterances mid-stream, taking into account the responses, or more often the lack of them, from recipients . . . As a result, what is produced is actually a joint production, which can hardly correspond to the speaker’s own initial intention or goal.” (Goodwin 1979; 1981) * communication accomplished via a trial-and-error process (Arundale 2008)

◮ Ginzburg et al: Dialogue understanding involves

modelling distinct fragment-categories via assigning distinct syntactic/semantic types (Ginzburg 2009)

◮ Dynamic Syntax: Dialogue understanding via

incremental interdependence of parsing/production

• speaker and hearer joint construction of turns via own context

Gregoromichelaki et al Osnabr¨ uck 19/09/09 28/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Outline

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction Dynamic Syntax: incremental structure/content growth Grammar design

Gregoromichelaki et al Osnabr¨ uck 19/09/09 29/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

The Process of Building up interpretation

◮ Building representations of content as goal-driven

monotonic tree-growth from word-sequence

Gregoromichelaki et al Osnabr¨ uck 19/09/09 30/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

The Process of Building up interpretation

◮ Building representations of content as goal-driven

monotonic tree-growth from word-sequence

• Parsing ‘Mary, John upset’

?Ty(t), ♦ →

Upset′(Mary ′)(John′), Ty(t), ♦ John′ Ty(e) Upset′(Mary ′), Ty(e → t) Mary ′ Ty(e) Upset′ Ty(e → (e → t))

NPs map onto (epsilon) terms of type e, propositions are of type t. All terms are concepts, induced from procedures given by words Scope evaluation defined on resulting tree.

Gregoromichelaki et al Osnabr¨ uck 19/09/09 30/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Actions all the way

◮ words as packages of actions, e.g.

Gregoromichelaki et al Osnabr¨ uck 19/09/09 31/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Actions all the way

◮ words as packages of actions, e.g.

• verbs induce (partial) propositional templates:

Gregoromichelaki et al Osnabr¨ uck 19/09/09 31/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Actions all the way

◮ words as packages of actions, e.g.

• verbs induce (partial) propositional templates:

upset IF {?Ty(e → t)} THEN make(↓1);go(↓); put(Fo(Upset′), Ty(e → (e → t))) go(↑1); make(↓0); go(↓0); put(?Ty(e)) ELSE ABORT

?Ty(e → t) ?Ty(e), ♦ Upset′

◮ Requirements, ?X drive all tree growth

Gregoromichelaki et al Osnabr¨ uck 19/09/09 31/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Underspecification: structural

◮ Processing non-contiguous dependencies

◮ e.g. ‘Mary, John upset’

Tn(0), ?Ty(t), ♦

Gregoromichelaki et al Osnabr¨ uck 19/09/09 32/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Underspecification: structural

◮ Processing non-contiguous dependencies

◮ e.g. ‘Mary, John upset’

‘Mary

Tn(0), ?Ty(t) Mary ′, ↑∗Tn(0) ?∃xTn(x), ♦

Gregoromichelaki et al Osnabr¨ uck 19/09/09 32/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Underspecification: structural

◮ Processing non-contiguous dependencies

◮ e.g. ‘Mary, John upset’

‘Mary

Tn(0), ?Ty(t) Mary ′, ↑∗Tn(0) ?∃xTn(x) ?Ty(e) ♦ ?Ty(e → t)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 32/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Underspecification: structural

◮ Processing non-contiguous dependencies

◮ e.g. ‘Mary, John upset’

‘Mary, John

Tn(0), ?Ty(t) Mary ′, ↑∗Tn(0) ?∃xTn(x) ?Ty(e), John′ ♦, Ty(e) ?Ty(e → t)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 32/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Underspecification: structural

◮ Processing non-contiguous dependencies

◮ e.g. ‘Mary, John upset’

‘Mary, John

Tn(0), ?Ty(t) Mary ′, ↑∗Tn(0) ?∃xTn(x) Ty(e), John′ ?Ty(e → t), ♦

Gregoromichelaki et al Osnabr¨ uck 19/09/09 32/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Underspecification: structural

◮ Processing non-contiguous dependencies

◮ e.g. ‘Mary, John upset’

‘Mary, John upset’

Tn(0), ?Ty(t) Mary ′, ↑∗Tn(0) ?∃xTn(x) Ty(e), John′ ?Ty(e → t) ?Ty(e) ♦ Upset′

Gregoromichelaki et al Osnabr¨ uck 19/09/09 32/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Underspecification: structural

◮ Processing non-contiguous dependencies

◮ e.g. ‘Mary, John upset’

‘Mary, John upset’

Tn(0), ?Ty(t) Mary ′, ↑∗Tn(0) ?∃xTn(x) Ty(e), John′ ?Ty(e → t) ?Ty(e) ♦ Upset′

merge

Gregoromichelaki et al Osnabr¨ uck 19/09/09 32/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Underspecification: structural

◮ Processing non-contiguous dependencies

◮ e.g. ‘Mary, John upset’

‘Mary, John upset’

Tn(0), ?Ty(t) Ty(e), John′ ?Ty(e → t), ♦ Ty(e), Mary ′ Upset′

Gregoromichelaki et al Osnabr¨ uck 19/09/09 32/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Underspecification: structural

◮ Processing non-contiguous dependencies

◮ e.g. ‘Mary, John upset’

‘Mary, John upset’

Tn(0), Ty(t), Upset′(Mary ′)(John′), ♦ Ty(e), John′ Ty(e → t), Upset′(Mary ′) Ty(e), Mary ′ Upset′

Gregoromichelaki et al Osnabr¨ uck 19/09/09 32/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Underspecification: content

◮ Pronouns project meta-variables (U)

Substituted by item from context during construction (1) Someone was smoking He fainted.

Gregoromichelaki et al Osnabr¨ uck 19/09/09 33/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Underspecification: content

◮ Pronouns project meta-variables (U)

Substituted by item from context during construction (1) Someone was smoking He fainted. Tree as Context: Tree under Construction:

Smoking ′(ǫ, x, Person′(x)) ǫ, x, Person′(x) Smoking ′ ?Ty(t) U, ?∃xFo(x) ♦ Faint′ substitution

Gregoromichelaki et al Osnabr¨ uck 19/09/09 33/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Context: parsing

◮ Context consists of a store of parse states, hence a set of

triples T, W , A (a (partial) tree, the string so far processed, the sequence of actions used)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 34/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Generation

◮ Speakers go through the same tree-growth actions, except they also

have a somewhat richer goal tree.

◮ Each word licensed must update partial tree towards the goal tree

• Generating Mary John upset

goal tree test tree

Upset′(Mary ′)(John′)Ty(t), ♦ John′ Ty(e) Upset′(Mary ′), Ty(e → t) Mary ′ Ty(e) Upset′ Ty(e → (e → t)) ?Ty(t), Tn(0), ♦

Gregoromichelaki et al Osnabr¨ uck 19/09/09 35/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Generation

◮ Speakers go through the same tree-growth actions, except they also

have a somewhat richer goal tree.

◮ Each word licensed must update partial tree towards the goal tree

• Generating Mary John upset

goal tree test tree

Upset′(Mary ′)(John′)Ty(t), ♦ John′ Ty(e) Upset′(Mary ′), Ty(e → t) Mary ′ Ty(e) Upset′ Ty(e → (e → t)) ?Ty(t), Tn(0) Mary ′, ↑∗Tn(0) ?∃xTn(x)♦

Gregoromichelaki et al Osnabr¨ uck 19/09/09 35/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Generation

◮ Speakers go through the same tree-growth actions, except they also

have a somewhat richer goal tree.

◮ Each word licensed must update partial tree towards the goal tree

• Generating Mary John upset

goal tree test tree

Upset′(Mary ′)(John′)Ty(t), ♦ John′ Ty(e) Upset′(Mary ′), Ty(e → t) Mary ′ Ty(e) Upset′ Ty(e → (e → t)) ?Ty(t), Tn(0) Mary ′, ↑∗Tn(0) ?∃xTn(x) ?Ty(e) John′ ♦ ?Ty(e → t)

Gen: ‘Mary

Gregoromichelaki et al Osnabr¨ uck 19/09/09 35/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Generation

◮ Speakers go through the same tree-growth actions, except they also

have a somewhat richer goal tree.

◮ Each word licensed must update partial tree towards the goal tree

• Generating Mary John upset

goal tree test tree

Upset′(Mary ′)(John′)Ty(t), ♦ John′ Ty(e) Upset′(Mary ′), Ty(e → t) Mary ′ Ty(e) Upset′ Ty(e → (e → t)) ?Ty(t), Tn(0) Mary ′, ↑∗Tn(0) ?∃xTn(x) ?Ty(e) John′ ♦, Ty(e) ?Ty(e → t)

Gen: ‘Mary John

Gregoromichelaki et al Osnabr¨ uck 19/09/09 35/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Generation

◮ Speakers go through the same tree-growth actions, except they also

have a somewhat richer goal tree.

◮ Each word licensed must update partial tree towards the goal tree

• Generating Mary John upset

goal tree test tree

Upset′(Mary ′)(John′)Ty(t), ♦ John′ Ty(e) Upset′(Mary ′), Ty(e → t) Mary ′ Ty(e) Upset′ Ty(e → (e → t)) ?Ty(t), Tn(0) Mary ′, ↑∗Tn(0) ?∃xTn(x) Ty(e) John′ ?Ty(e → t), ♦

Gen: ‘Mary John

Gregoromichelaki et al Osnabr¨ uck 19/09/09 35/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Generation

◮ Speakers go through the same tree-growth actions, except they also

have a somewhat richer goal tree.

◮ Each word licensed must update partial tree towards the goal tree

• Generating Mary John upset

goal tree test tree

Upset′(Mary ′)(John′)Ty(t), ♦ John′ Ty(e) Upset′(Mary ′), Ty(e → t) Mary ′ Ty(e) Upset′ Ty(e → (e → t)) ?Ty(t), Tn(0) Mary ′, ↑∗Tn(0) ?∃xTn(x) Ty(e) John′ ?Ty(e → t) ?Ty(e) ♦ Upset′

Gen: ‘Mary John upset

Gregoromichelaki et al Osnabr¨ uck 19/09/09 35/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Generation

◮ Speakers go through the same tree-growth actions, except they also

have a somewhat richer goal tree.

◮ Each word licensed must update partial tree towards the goal tree

• Generating Mary John upset

goal tree test tree

Upset′(Mary ′)(John′)Ty(t), ♦ John′ Ty(e) Upset′(Mary ′), Ty(e → t) Mary ′ Ty(e) Upset′ Ty(e → (e → t)) ?Ty(t), Tn(0) Mary ′, ↑∗Tn(0) ?∃xTn(x) Ty(e) John′ ?Ty(e → t) ?Ty(e) ♦ Upset′ merge

Gen: ‘Mary John upset

Gregoromichelaki et al Osnabr¨ uck 19/09/09 35/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Generation

◮ Speakers go through the same tree-growth actions, except they also

have a somewhat richer goal tree.

◮ Each word licensed must update partial tree towards the goal tree

• Generating Mary John upset

goal tree test tree

Upset′(Mary ′)(John′)Ty(t), ♦ John′ Ty(e) Upset′(Mary ′), Ty(e → t) Mary ′ Ty(e) Upset′ Ty(e → (e → t)) ?Ty(t), Tn(0) Ty(e) John′ ?Ty(e → t), ♦ Ty(e), Mary ′ Upset′

Gen: ‘Mary John upset

Gregoromichelaki et al Osnabr¨ uck 19/09/09 35/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Context: generation

◮ Generation thus characterised in exactly same terms as

parsing, except in current parse state, current partial tree must subsume, ⊒, the goal tree.

◮ Context uniformly defined in parsing and generation: a set of

parse states: (partial) tree structure, (partial) string, sequence

• f actions.

Gregoromichelaki et al Osnabr¨ uck 19/09/09 36/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Ellipsis - Filling out interpretation from Context

◮ With semantics as structural representations of content,

syntax as the process of constructing such representations, production and parsing as both using the same processes

◮ Context can be defined as a store of evolving structures +

actions used to build them Purver et al 2007, Cann et al 2009

◮ Users can retrieve actions stored in context and re-use those

to build up interpretation

Gregoromichelaki et al Osnabr¨ uck 19/09/09 37/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Ellipsis - Filling out interpretation from Context

◮ With semantics as structural representations of content,

syntax as the process of constructing such representations, production and parsing as both using the same processes

◮ Context can be defined as a store of evolving structures +

actions used to build them Purver et al 2007, Cann et al 2009

◮ Users can retrieve actions stored in context and re-use those

to build up interpretation

◮ The parallelism effects in ellipsis (structural and semantic)

follow immediately

Gregoromichelaki et al Osnabr¨ uck 19/09/09 37/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Ellipsis (a): re-use of structure

◮ Using structure from context - parser/generator starts from

partial tree: (2) Q: Who did John upset ? Ans: Himself. Context Tree: becomes Tree under Construction:

Upset′(WH)(John′) John′ Upset′(WH) WH Upset′ Upset′(WH)(John′) John′ Upset′(WH) WH, Ty(e), ♦ Upset′

substitution (3) Q: Who did every woman ignore? Ans: Her husband.

Gregoromichelaki et al Osnabr¨ uck 19/09/09 38/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Ellipsis (b): re-using actions from context

Interpreting: (4) Who hit himself? John did. Context Tree under Construction

Upset′(WH)(WH) WH Upset′(WH) WH Upset′ ?Ty(t) John′ U, Ty(e → t) ♦

context actions to re-run actions of upset actions of reflexive completing/evaluating tree

Gregoromichelaki et al Osnabr¨ uck 19/09/09 39/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Ellipsis & Syntactic constraints – ACE

◮ Antecedent-contained ellipsis constraints emerge from encoded

incremental growth (5) Bill saw someone [ that John did ]

Gregoromichelaki et al Osnabr¨ uck 19/09/09 40/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Ellipsis & Syntactic constraints – ACE

◮ Antecedent-contained ellipsis constraints emerge from encoded

incremental growth (5) Bill saw someone [ that John did ] Tn(0), ?Ty(t) Bill′ ?Ty(e → t) ?Ty(e) ?Ty(cn) x, Ty(e) Person′ λP(ǫ, x, P(x)) See′ Bill saw someone

Gregoromichelaki et al Osnabr¨ uck 19/09/09 40/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Ellipsis & Syntactic constraints – ACE

◮ Antecedent-contained ellipsis constraints emerge from encoded

incremental growth (5) Bill saw someone [ that John did ] Tn(0), ?Ty(t) Bill′ ?Ty(e → t) ?Ty(e) ?Ty(cn) x, Ty(e) Person′ λP(ǫ, x, P(x)) See′ ?Ty(t) x L Bill saw someone that

Gregoromichelaki et al Osnabr¨ uck 19/09/09 40/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Ellipsis & Syntactic constraints – ACE

◮ Antecedent-contained ellipsis constraints emerge from encoded

incremental growth (5) Bill saw someone [ that John did ] Tn(0), ?Ty(t) Bill′ ?Ty(e → t) ?Ty(e) ?Ty(cn) x, Ty(e) Person′ λP(ǫ, x, P(x)) See′ ?Ty(t) x John′ U L Bill saw someone that John did

Gregoromichelaki et al Osnabr¨ uck 19/09/09 40/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Ellipsis & Syntactic constraints – ACE

◮ Antecedent-contained ellipsis constraints emerge from encoded

incremental growth (5) Bill saw someone [ that John did ] Tn(0), ?Ty(t) Bill′ ?Ty(e → t) ?Ty(e) ?Ty(cn) x, Ty(e) Person′ λP(ǫ, x, P(x)) See′ ?Ty(t) x John′ U ?Ty(e)See′ L Bill saw someone that John did re-run: see

Gregoromichelaki et al Osnabr¨ uck 19/09/09 40/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Ellipsis & Syntactic constraints – ACE

◮ Antecedent-contained ellipsis constraints emerge from encoded

incremental growth (5) Bill saw someone [ that John did ] Tn(0), ?Ty(t) Bill′ ?Ty(e → t) ?Ty(e) ?Ty(cn) x, Ty(e) Person′ λP(ǫ, x, P(x)) See′ ?Ty(t) John′ U x See′ L Bill saw someone that John did re-run: see unification

Gregoromichelaki et al Osnabr¨ uck 19/09/09 40/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Ellipsis & Syntactic constraints – ACE

◮ Antecedent-contained ellipsis constraints emerge from encoded

incremental growth (5) Bill saw someone [ that John did ] Tn(0), ?Ty(t) Bill′ ?Ty(e → t) ?Ty(e) ?Ty(cn) x, Ty(e) Person′ λP(ǫ, x, P(x)) See′ ?Ty(t) John′ U x See′ L Bill saw someone that John did re-run: see unification completion of tree:

Gregoromichelaki et al Osnabr¨ uck 19/09/09 40/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

ACE update in main tree

(5) Bill saw someone that John did Ty(t) See′( (ǫ, x, Person′(x) ∧ See′(x)(John)) (Bill′) ) Bill′ Ty(e → t) Ty(e) ǫ, x, Person′(x) ∧ See′(x)(John) Ty(cn) x, Person′(x) ∧ See′(x)(John′) x, Ty(e) Person′ λP(ǫ, x, P(x)) See′ Ty(t), See′(x)(John′) John′ See′(x) U x See′ L

Gregoromichelaki et al Osnabr¨ uck 19/09/09 41/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Ellipsis: re-use of actions induces parallellism

◮ Using actions from context – sloppy readings:

(6) John upset his mother. Harry too. (7) The man [who arrested John] failed to read him his rights. (8) The man who arrested Tom did too.

◮ Also more general parallellism effects: scope, construction

type.... (9) A man, I certainly wouldn’t appoint. A friend of mine, I just might.

Gregoromichelaki et al Osnabr¨ uck 19/09/09 42/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Dynamic Syntax: Why suited for split utterance analysis?

◮ It’s action-based: “syntax” defined as actions that determine

how interpretation is built up online

◮ It’s goal-directed/strongly predictive: anticipation of goals,

propositional and sub-propositional

◮ It’s incremental and context-dependent: proceeds

left-to-right/word-by-word, with context-based interpretation update at every input

◮ It’s bi-directional: the same set of mechanisms used both for

parsing and generation (predicting speaker/hearer switch)

◮ Generation and parsing equally context-dependent

Gregoromichelaki et al Osnabr¨ uck 19/09/09 43/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Split utterances

◮ Partial tree as context/input for speaker and hearer

(10) A: John saw ... B: Bill. context: test/parse tree: goal tree:

?Ty(t) John′ ?Ty(e → t) ?Ty(e) ♦ See′ ?Ty(t) John′ ?Ty(e → t) ?Ty(e) See′ Ty(t), See′(Bill′)(John′) John′ Ty(e → t), See′(Bill′) Bill′ See′

Gregoromichelaki et al Osnabr¨ uck 19/09/09 44/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Split utterances

◮ Partial tree as context/input for speaker and hearer

(10) A: John saw ... B: Bill. context: test/parse tree: goal tree:

?Ty(t) John′ ?Ty(e → t) ?Ty(e) ♦ See′ ?Ty(t) John′ ?Ty(e → t) Bill,♦ See′ Ty(t), See′(Bill′)(John′) John′ Ty(e → t), See′(Bill′) Bill′ See′ * Testing and producing Bill

Gregoromichelaki et al Osnabr¨ uck 19/09/09 44/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Split utterances

◮ Partial tree as context/input for speaker and hearer

(10) A: John saw ... B: Bill. context: test/parse tree: goal tree:

?Ty(t) John′ ?Ty(e → t) ?Ty(e) ♦ See′ ?Ty(t) John′ ?Ty(e → t), ♦ Bill See′ Ty(t), See′(Bill′)(John′) John′ Ty(e → t), See′(Bill′) Bill′ See′ * Testing and producing Bill * Pointer movement

Gregoromichelaki et al Osnabr¨ uck 19/09/09 44/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Split utterances

◮ Partial tree as context/input for speaker and hearer

(10) A: John saw ... B: Bill. context: test/parse tree: goal tree:

?Ty(t) John′ ?Ty(e → t) ?Ty(e) ♦ See′ Ty(t), See′(Bill′)(John′), ♦ John′ See′Bill′ Bill See′ Ty(t), See′(Bill′)(John′) John′ Ty(e → t), See′(Bill′) Bill′ See′ * Testing and producing Bill * Pointer movement * Completion

Gregoromichelaki et al Osnabr¨ uck 19/09/09 44/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Ellipsis: a unitary account

◮ With context as reflecting terms (content), structure and

actions

◮ The account of ellipsis reflects directly the folk intuition ◮ Ellipsis parallels anaphora as intrinsically context-dependent ◮ Heterogeneity of resulting content captured ◮ Syntactic restrictions on ellipsis types expressible through tree

growth dynamics

Gregoromichelaki et al Osnabr¨ uck 19/09/09 45/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Outline

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dialogue data What does context-dependence amount to? Language as action alternative: understanding via interaction Dynamic Syntax: incremental structure/content growth Grammar design

Gregoromichelaki et al Osnabr¨ uck 19/09/09 46/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Dialogue and Ellipsis processing

◮ Implications for processing complexity

• economy in phonological length/articulatory effort
• coherence device (as anaphora in general)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 47/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Dialogue and Ellipsis processing

◮ Implications for processing complexity

• economy in phonological length/articulatory effort
• coherence device (as anaphora in general)

◮ DS adds economy in repeating parsing/generation choices

to resolve underspecification parallelism/alignment preserves types of choice made in the antecedent

Gregoromichelaki et al Osnabr¨ uck 19/09/09 47/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Fragments as interruptions (a)

◮ Bonus to hearer of interruptive use of fragments in dialogue

• Immediate clarification identifies what is being queried,

removing what otherwise yields structural ambiguity for hearer

• reduces interpretational ambiguity as to who is referred to,

by resolution through interaction

Gregoromichelaki et al Osnabr¨ uck 19/09/09 48/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Fragments as interruptions (a)

◮ Bonus to hearer of interruptive use of fragments in dialogue

• Immediate clarification identifies what is being queried,

removing what otherwise yields structural ambiguity for hearer

• reduces interpretational ambiguity as to who is referred to,

by resolution through interaction

• allows joint construction of meanings (split utterances)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 48/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Fragments as interruptions (a)

◮ Bonus to hearer of interruptive use of fragments in dialogue

• Immediate clarification identifies what is being queried,

removing what otherwise yields structural ambiguity for hearer

• reduces interpretational ambiguity as to who is referred to,

by resolution through interaction

• allows joint construction of meanings (split utterances)

◮ These economies critically make use of incrementality

as provided by the grammar

Gregoromichelaki et al Osnabr¨ uck 19/09/09 48/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Fragments as interruptions: (b)

◮ partiality of goal trees benefits both speakers in dialogue:

• incrementality in production (Levelt 1989)
• “speakers recraft their utterances mid-stream, taking into

account the responses, or more often the lack of them, from recipients . . . As a result, what is produced is actually a joint production, which can hardly correspond to the speaker’s own initial intention or goal.” (Goodwin 1979; 1981)

speakers can start out with only partial thought in mind speakers can intervene with some partial contribution to this emergent structure

◮ Joint construction of turns emerging as the product of a

dynamic process of interaction grounded on parsing/production interdependence

Gregoromichelaki et al Osnabr¨ uck 19/09/09 49/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Consequences for grammar/pragmatics interface

◮ Incrementality, predictivity and partiality in

parsing/production minimizes need for “guessing” speaker intentions in modelling split utterances

◮ hence

role of mutual knowledge/common ground computation minimized

Gregoromichelaki et al Osnabr¨ uck 19/09/09 50/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Consequences for grammar/pragmatics interface

◮ Incrementality, predictivity and partiality in

parsing/production minimizes need for “guessing” speaker intentions in modelling split utterances

◮ hence

role of mutual knowledge/common ground computation minimized

◮ but success in communication will rely on

incremental correction/clarification and sequential structure of dialogue (Arundale 2008)

Gregoromichelaki et al Osnabr¨ uck 19/09/09 50/62

Linguistic knowledge: the relevance of dialogue Dialogue Modelling: new challenge, new data, new solutions Dynamic Syntax: incremental structure/content growth Grammar design

Grammars reflecting processing as linguistic knowledge

◮ Language is a tool-box for constructing formal languages, a

system with semantic flexibility in perpetual flux: Cooper and

Ranta 2008, Larsson 2008

◮ Syntax as growth of semantic representations: Cann et al 2005,

Purver et al 2006

◮ Constraint-based grammars as mechanisms for utterance

processing reflecting incrementality of parsing and production (cp categorial grammars: eg Steedman 1996, 2000) Hawkins

1994, 2004, Kempson et al 2001

◮ Grammar as mechanisms for (conversational)

interaction:Gregoromichelaki et al 2009

Gregoromichelaki et al Osnabr¨ uck 19/09/09 51/62