=Paper=
{{Paper
|id=Vol-1863/paper_13
|storemode=property
|title=Procedural syntax and interactions
|pdfUrl=https://ceur-ws.org/Vol-1863/paper_13.pdf
|volume=Vol-1863
|authors=Eleni Gregoromichelaki
|dblpUrl=https://dblp.org/rec/conf/esslli/Gregoromichelaki17
}}
==Procedural syntax and interactions==
https://ceur-ws.org/Vol-1863/paper_13.pdf
Procedural Syntax and Interactions
Eleni Gregoromichelaki
King’s College London
University of Osnabrueck
elenigregor@gmail.com
Abstract interleaved and interacting with other activities and
The view of NLs as codes mediating a map- agents. Accordingly, in everyday conversation, ut-
ping between “expressions” and the world is terances are not expected to display evidence of
abandoned to give way to a view where utter- necessary hierarchical constituency, e.g. senten-
ances are seen as actions aimed to locally and tial structuring: non-sentential utterances are ade-
incrementally alter the affordances of the con- quate to underpin interlocutor coordination and all
text. Such actions employ perceptual stimuli linguistic dependencies are resolvable across more
composed not only of “words” and “syntax” than one turn:
but also elements like visual marks, gestures, (1) Angus: But Domenica Cyril is an intelligent
sounds, etc. Any such stimuli can participate and entirely well-behaved dog who
in the domain-general processes that consti- Domenica: happens to smell
tute the “grammar”. The function of the gram- [radio play, 44 Scotland Street]
mar is dynamic categorisation of various per- In such cases, postulating a notion of well-
ceptual inputs and their integration in the pro- formedness based on a code licensing units rang-
cess of generating the next action steps. Given ing over strings of words, as an independent level
these assumptions, a challenge that arises is of structuring, impedes a natural account of such
how to account for the reification of such pro- phenomena. This is because joining overt forms to-
cesses as exemplified in apparent metarepre- gether often results in illformedness or misleading
sentational practices like quotation, reporting, interpretations:
citation etc. It is argued that even such phe-
nomena can receive adequate and natural ex- (2) A: I heard a bang. Did you hurt
planations through a grammar that allows for B: myself? No, but Mary is in a state
the ad hoc creation of occasion-specific con- Moreover, at the level of semantics/pragmatics of
tent through reflexive mechanisms. dialogue, the issue of recoverability of proposi-
tional intentions is also problematic, e.g., in cases
1 Language as action and grammar such as (5) where various speech acts are accom-
Standard models that describe natural languages plished within the unfolding of a shared single
(NLs) as representational systems belong to the proposition (see Gregoromichelaki et al. (2011)):
‘language-as-product’ paradigm (Clark, 1992), (3) Jack: I just returned
concerned with the definition of linguistic repre- Kathy: from . . .
sentations, the “product” of linguistic processing. Jack: Finland. [Lerner (2004)]
In this tradition, it has been a standard assumption (4) Eleni: A: Are you left or
that NL properties should be explained by reifying Yo: Right-handed. [natural data]
NLs as abstract codes, mapping forms (strings of (5) Hester Collyer: It’s for me.
symbols) to propositional intentions. However, a Mrs Elton the landlady: And Mr. Page?
substantial amount of evidence indicates that NL Hester Collyer: is not my husband. But I’d
use substantially affects NL structuring indicating rather you think of me as Mrs. Page. [The
an alternative characterisation: within a ‘language Deep Blue Sea (film)]
as action’ paradigm, NL properties can be expli-
cated as coinciding with those of human action; This endemic context-sensitivity and situated-
an agent’s linguistic actions are structured sequen- ness of NL use is indicative of the fact that both
tially, directed by predictions of upcoming inputs, content and structure are emergent products of the
processes and practices underpinning human inter- for the resolution of anaphora and ellipsis, espe-
action. For these reasons, the more general ap- cially “sloppy” or “paycheck” readings, whose res-
proach to NL analysis argued for here revolves olution relies on re-executing (‘rerunning’) previ-
around the idea that structures, objects, concepts, ous action sequences in an updated processing en-
concrete reality (and even the individual self) can vironment. Maintaining abandoned options is re-
all be taken as metaphysically emergent categories quired for the modelling of backtracking in sub-
with processes, mechanisms, and change as onto- sententially occurring conversational phenomena
logically primary.1 like clarification, self-/other-corrections, etc. but
also humour effects and puns (Gregoromichelaki,
2 DS-TTR in press). Consequently, coordination among inter-
A grammar architecture adopting this perspective locutors is seen not as inferential metarepresenta-
can be articulated within DS-TTR (Cann et al. tional activity but as the outcome of the fact that
(2005); Purver et al. (2010); Gregoromichelaki the grammar consists of a set of licensed comple-
(in press)). Here NLs are conceived as compris- mentary actions that both speakers and hearers have
ing sets of processes modelled formally as proce- to perform in synchrony (Gregoromichelaki et al.,
dures. Both NLs’ temporal structuring (syntax) 2013).
and lexical specifications are analysed as involv-
ing stored sequences (macros) of elementary (epis- 2.1 Quotation in DS-TTR
temic) actions, defined in an IF-THEN-ELSE for- Given these assumptions, a challenge that arises
mat. Such actions incrementally and predictively is how to account for the reification of grammati-
build or linearise conceptual categories expressed cal processes as exemplified in apparent metarep-
in TTR-representations (Cooper, 2012). The model resentational practices like quotation, reporting, ci-
assumes tight interlinking of NL perception and tation etc. As we saw earlier in (1)-(5), perfectly
action: production uses simulation and testing of intelligible moves in dialogue can be achieved sim-
parse states in order to license the generation of ply by initiating a grammatical dependency which
strings; comprehension predictively builds struc- prompts either interlocutor to fulfill it without spe-
tures to accommodate upcoming inputs in order to cific determination or identifiability of a given
constrain efficiently the usual overwhelming ambi- speech-act. In various other cases though, the in-
guity of NL stimuli. By imposing top-down predic- terlocutor completing somebody else’s utterance
tive and goal-directed processing at all comprehen- might be seen as offering the completion along with
sion and production stages, interlocutor feedback a query as to whether such a (meta)representation
is incrementally anticipated and integrated. The is what the other interlocutor would have said (e.g.
model includes subsentential tracking of the shift- (2)). There are further such phenomena in cases of
ing contextual parameters of each word-utterance citation, quotation, reports, echoing uses, and code-
event (Eshghi et al. (2015); Gregoromichelaki (in switching:
press)). Context constitutes an integral part of the (6) “Cities,” he said, “are a very high priority.”
grammar, not only as a record of the shifting pa- (7) Wright won’t disclose how much the Nike
rameters that provide for the interpretation of vari- deal is worth, saying only that “they treat me
ous indexical elements (e.g. myself in (2)), but also well”. [De Brabanter (2010)]
storing (a) the emergent (partial) structures con-
(8) A doctor tells him [Gustave Flaubert] he is
structed from the contributions of all participants;
like a “vieille femme hysterique”; he agrees.
(b) the phonological/graphical elements that have
[De Brabanter (2010)]
been employed; (c) the actions used, recorded as
(9) Alice said that life is “difficult to
traversals of paths in a graph display; (d) processing
understand”. [Cappelen and Lepore (1997)]
paths that have been considered as probabilistically
(10) Mary felt relieved. If Peter came tomorrow,
live options but not eventually pursued (Sato, 2011;
she would be saved. [Recanati (2010)]
Hough, 2015). Storing the action paths is necessary
1
Despite recent attempts to integrate such phenom-
This view has its roots in an ancient philosophical pro-
gramme starting in the Western world with Heraclitus, situated
ena within standard grammars (e.g., (Ginzburg and
within a tradition following, among others, Martin Heidegger, Cooper, 2014; Maier, 2014; Potts, 2007)), certain
Ilya Prigogine, Gilles Deleuze, and even encompassing cur- data are not amenable to appropriate treatment due
rent notions like the concept of the extended mind (Clark and
Chalmers, 1998; Clark, 2008). to the lack of modelling incrementality within these
formalisms. For example, as can be seen in (6)-(9), A DS-TTR grammar takes words (and the oper-
quotation can appear subsententially, and discon- ation of “syntax” in general) as offering affor-
tinuously, at any point, which means that contextual dances exploited by the interlocutors to facilitate
parameters regarding the utterance event and se- interaction. This means that words and linguistic
mantic evaluation need to be able to shift incremen- constructions are NOT conceptualised as abstract
tally at each word-by-word processing stage. Addi- code elements, expression types, that are associated
tionally, quotation is one of the environments where with referential/semantic values (cf Cooper (2014)
the phenomenon of split-utterances is observed fre- where string structure is still presumed). With no
quently as an opportunity arises for co-constructing privileged semantic entities corresponding to (types
a vivid unified perspective of some (actual or imag- of) expressions, only domain-general mechanisms
inary) speech/thought event (Gregoromichelaki, in for processing stimuli, quotation thus offers a cru-
press): cial test for the legitimacy of these DS-TTR claims:
(11) Clinician: So I watch this person being when processing a quoted/cited string, what hap-
killed and then I go to bed and I’m you know pens within the quotation marks (or any other indi-
lying there going, “well” cations) according to these assumptions?
Patient: “did I hear something?” [Duff et al. In fact, it turns out that such cases are also
(2007)] unproblematic for the DS-TTR model, and can
The contextual parameters relevant to the resolu- be explicated in a natural manner that conforms
tion of indexicals (e.g. I) in such cases, even though with intuitions and parallels the modelling of
needing to shift mid-sentence, do not necessarily anaphora/ellipsis. First, in order to model cases
track the current speaker/hearer roles. Moreover, like (6)-(10), (14), (15), as well as mid-sentence
such role-switches include cases where the same general code-switching, it has to be assumed that
structure can be employed both as expressing a the context keeps track incrementally, through a
speaker’s own voice and as a subsequent quotation: designated metavariable (g in (16)), of which and
(12) A: SOMEONE is keen [BBC] whose grammar is being employed at each partic-
B: says the man who slept here all night ular subsentential stage (cf Ginzburg and Cooper
(2014)). Next, consider the most challenging cases,
In all such cases, issues of “footing” (Goffman, namely, metalinguistic uses, for example (13), so-
1979), namely changes in perspectives and roles called pure quotation, where an NL-string appears
assumed by interlocutors, intersect with syntac- in a regular NP position. Under DS-TTR assump-
tic/semantic issues of direct/indirect speech con- tions, this will be a pointer position where the
structions and speech-act responsibility and echo- grammar has already generated a prediction for
ing. For these reasons, an adequate account of the processing of a singular term (?T y(e), other
the function of such NL devices can be given cases might involve ?T y(cn), etc.). The expla-
straighforwardly in DS-TTR due to its incremen- nation of what happens here is based on the fact
tal modelling of context shifting, the potential for that actions are first-class citizens in DS-TTR. This
sharing of syntactic/semantic dependencies, and means that previous actions can be invoked by the
the fact that there is no requirement to derive a grammar to be re-executed (‘rerun’) in order to
global propositional speech act (Gregoromichelaki provide parallel but distinct contents as needed in
(in press); Gregoromichelaki & Kempson (2016)). cases of sloppy-ellipsis or paycheck-pronoun read-
On the other hand, modelling the potential of ings. From this perspective, metalinguistic, echoic,
partially assuming another speaker’s role, being and similar uses are cases where the actions spec-
perceived as “demonstrating” what somebody else ified by some grammar g for processing a partic-
was going to say, and the “metalinguistic” appear- ular string, e.g. the embedded sentential string in
ance of various such phenomena might seem espe- (13), come to be executed on the spot to provide
cially problematic aspects for the DS-TTR stance: an ad hoc conceptualisation of a demonstrated ac-
(13) “Life is difficult” is grammatical. tion. The formalisation of the basic mechanism
(14) James says that “Quine” wants to speak to us. is shown in (16) below. Different variants of this
[James thinks that McPherson is Quine] macro and combinations with other independently
(15) “I talk better English than the both of needed components of the grammar account for all
youse!” shouted Charles, thereby convincing such phenomena:
me that he didn’t.
(18) John saw the spider and was like “ahh!”
(19) John was eating like [gobbling gesture]
(20) She went “Mm Mmmrn Mphh”
The existence of such compositions, along with all
(16)
(a) demonstration action the previous data, might be challenging, under one
In (16), the higher-order action run, also employed construal, for the account of NL-gesture coordi-
in cases of sloppy anaphora, is triggered. run is nation in Rieser (2014; 2015). Rieser presents a
parameterised to some grammar g (replacing the framework (the λ-π calculus) where NL and ges-
metavariable g), which can be distinct from the ture are modelled as independent but communi-
grammar used for parsing/producing the rest of the cating processes. Even though the process meta-
string (see (8), (15)). At the same time, the ex- physics incidentally mentioned there is a wel-
ecuted sequence of actions hαi , ..., αn i, bound to come development, the assumption of indepen-
the rule-level variables hai ,...,an i, confers the ad dence might be questioned. First, this assump-
hoc conceptual type of the quoting utterance event tion seems to be an artifact of presupposing that
uq which therefore functions as a demonstration. NLs are structured codes mediating arbitrary map-
The performance of this demonstration event is pings from standard syntactic forms (trees inhab-
then categorised as belonging to the already pre- ited by words) to propositional meanings (e.g. λ-
dicted semantic type, here, in (13), a referential calculus formulae). Since the co-speech gestures
term (T y(e)) (feasible due to TTR’s subtyping def- examined are related to imagery (aural, visual, etc.)
initions). The rest of the string then delivers a con- in an iconic manner, modelling their contribution
tent that combines with the reification of this ad hoc in the standard way needs to abstract representa-
execution. In (13), this delivers the interpretive re- tions from the kinematics that cannot be unified
sult that this demonstration of the execution of the with NL syntactic representations. In contrast, the
grammatical actions is characterised as having the view taken here is that NL actions do not require
property derived from processing is grammatical. an independent syntax relying on the hierarchical
For echoic cases, where the interpretation of the structuring of stimuli sequences. Hence produc-
indexicals shifts following parameter values sup- tion/comprehension of stimuli in various modali-
plied by the invoked context, e.g. (7), (15), a sim- ties need not be segregated. Second, the major
ilarly triggered action execution is accomplished, argument in Rieser’s analysis comes from SaGA
this time, with parallel introduction of the quoting data (Lücking et al., 2013) where NL segments and
context as a mentioned utterance event u, replacing gesture-strokes seem not to synchronise perfectly.
the metavariable u of type es , i.e. eventuality: However, this is not an argument for considering
(17) such stimuli qualitatively distinct. Perfect synchro-
nisation is not necessary within a single modality
either, e.g. dialogue participants do not perfectly
synchronise their turns. In a predictive framework
like DS-TTR, such asynchrony might reveal a pur-
pose, for example, co-speech gestures can be mod-
(b) demonstration-and-echoing action elled as elaborating or narrowing down predictions
Cases of direct quotation (e.g. (11), (12), (15)) that precede the processing of NL input. But then,
are those where such a freely-available contextual under this view, there is a viable and useful applica-
switch has been grammaticalised in English. tion of the λ-π calculus in the DS-TTR framework
Notably, given that the DS-TTR grammar does too. Given that DS-TTR processing is strictly in-
not provide form-meaning correspondences but cremental pursuing only one path at a time, it is
only provides for the parsing/generation of stimuli possible that various sources of information might
in context, the same mechanism can be applied to compete for sequential positions. Introducing ad-
non-linguistic signals/demonstrations: reifying the hoc channel interfaces, modelled with resources
processing of some upcoming element to provide from the λ-π calculus, can provide for the imple-
ad hoc content of another already predicted type ex- mentation of a sequentiality-repair mechanism, or-
plains how non-linguistic signals can compose sub- dering inputs/outputs, even within the same modal-
sententially with linguistic ones as the conceptuali- ity, so that they can be processed strictly incremen-
sation of some experience being demonstrated: tally.
References E. Gregoromichelaki, R. Kempson, and C. Eshghi
Howes. 2013. On making syntax dynamic: the chal-
Ronnie Cann, Ruth Kempson, and Lutz Marten. 2005. lenge of compound utterances and the architecture of
The Dynamics of Language. Elsevier, Oxford. the grammar. In Ipke Wachsmuth, Jan de Ruiter, Pe-
Herman Cappelen and Ernie Lepore. 1997. Varieties of tra Jaecks, and Stefan Kopp, editors, Alignment in
quotation. Mind, 106(423):429–450. Communication: Towards a New Theory of Commu-
nication. John Benjamins, “Advances in Interaction
Andy Clark and David Chalmers. 1998. The extended Studies”.
mind. Analysis, pages 7–19.
Eleni Gregoromichelaki. in press. Quotation in di-
Herbert H. Clark. 1992. Arenas of Language Use. Uni- alogue. In & Johnson M. Saka, P., editor, The
versity of Chicago Press. Semantics and Pragmatics of Quotation. Dordrecht:
Springer.
Andy Clark. 2008. Supersizing the mind: Embodiment,
action, and cognitive extension. OUP USA. Julian Hough. 2015. Modelling Incremental Self-
Repair Processing in Dialogue. Ph.D. thesis, Queen
Robin Cooper. 2012. Type theory and semantics in flux. Mary University of London.
In R. Kempson, T. Fernando, and N. Asher, editors,
Handbook of the Philosophy of Linguistics, volume Gene H. Lerner. 2004. Collaborative turn sequences.
14: Philosophy of Linguistics, pages 271–323. Else- In Conversation analysis: Studies from the first gen-
vier. eration, pages 225–256. Gene H. Lerner, John Ben-
jamins.
Robin Cooper. 2014. Phrase structure rules as dialogue
update rules. In V. Rieser and P. Muller, editors, Pro- Andy Lücking, Kirsten Bergman, Florian Hahn, Stefan
ceedings of DialWatt - Semdial 2014: The 18th Work- Kopp, and Hannes Rieser. 2013. Data-based analysis
shop on the Semantics and Pragmatics of Dialogue, of speech and gesture: The Bielefeld Speech and Ges-
pages 4352, Edinburgh, 13 September 2014. ture Alignment Corpus (SaGA) and its applications.
Journal on Multimodal User Interfaces, 7(1-2):5–18.
Philippe De Brabanter. 2010. The semantics and prag-
matics of hybrid quotations. Language and Linguis- Emar Maier. 2014. Mixed quotation: The grammar of
tics Compass, 4(2):107–120. apparently transparent opacity. Semantics & prag-
matics, 7(7):1–67.
Melissa C Duff, Julie A Hengst, Daniel Tranel, and
Neal J Cohen. 2007. Talking across time: Using Christopher Potts. 2007. The dimensions of quotation.
reported speech as a communicative resource in am- In Chris Barker and Polly Jacobson, editors, Proceed-
nesia. Aphasiology, 21(6-8):702–716. ings from the workshop on direct compositionality,
pages 405–431. Oxford: Oxford University Press.
Arash Eshghi, Christine Howes, Eleni Gre-
goromichelaki, Julian Hough, and Matthew Purver. Matthew Purver, Eleni Gregoromichelaki, Wilfried
2015. Feedback in conversation as incremental Meyer-Viol, and Ronnie Cann. 2010. Splitting the
semantic update. In Proceedings of the 11th Inter- i’s and crossing the you’s: Context, speech acts and
national Conference on Computational Semantics, grammar. In P. Łupkowski and M. Purver, editors,
pages 261–271. Proceedings of SemDial 2010 (PozDial), pages 43–
50, Poznan, Poland, June 2010. Polish Society for
Jonathan Ginzburg and Robin Cooper. 2014. Quota- Cognitive Science.
tion via dialogical interaction. Journal of Logic, Lan-
guage and Information, 23(3):287–311. François Recanati. 2010. Truth-conditional pragmat-
ics. Clarendon Press Oxford.
Erving Goffman. 1979. Footing. Semiotica, 25(1-2):1–
30. Hannes Rieser. 2014. Gesture and speech as au-
tonomous communicating processes. MS, University
Eleni Gregoromichelaki and Ruth Kempson. 2016. of Bielefeld.
Reporting, dialogue, and the role of grammar. In
Alessandro Capone, Ferenc Kiefer, and Franco Lo Pi- Hannes Rieser. 2015. When hands talk to mouth. ges-
paro, editors, Indirect reports and pragmatics, pages ture and speech as autonomous communicating pro-
115–150. Springer. cesses. In Proceedings of SEMDIAL 2015-GoDIAL,
page 122.
Eleni Gregoromichelaki, Ruth Kempson, Matthew
Purver, Gregory J. Mills, Ronnie Cann, Wilfried Yo Sato. 2011. Local ambiguity, search strategies and
Meyer-Viol, and Patrick G. T. Healey. 2011. Incre- parsing in Dynamic Syntax. In R. Kempson, E. Gre-
mentality and intention-recognition in utterance pro- goromichelaki, and C. Howes, editors, The Dynamics
cessing. Dialogue and Discourse, 2(1):199–233. of Lexical Interfaces. CSLI Publications.