Grammars as Parsers: Meeting the Dialogue Challenge

Abstract

Standard grammar formalisms are defined without reflection of the incremental, serial and context-dependent nature of language processing; any incrementality must therefore be reflected by independently defined parsing and/or generation techniques, and context-dependence by separate pragmatic modules. This leads to a poor setup for modelling dialogue, with its rich speaker-hearer interaction and high proportion of contextdependent and apparently grammatically ill-formed utterances. Instead, this paper takes an inherently incremental grammar formalism, Dynamic Syntax (DS) , proposes a context-based extension and defines corresponding context-dependent parsing and generation models together with a resulting natural definition of context-dependent well-formedness. These are shown to allow a straightforward model of otherwise problematic dialogue phenomena such as shared utterances, ellipsis and alignment. We conclude that language competence is a capacity for dialogue.

Context and Well-Formedness: The Dynamics of Ellipsis

“…16 Generation, according to (Otsuka and Purver 2003;Purver et al 2006), 17 uses the same lexical entries and actions as parsing (these being constitutive of the grammar formalism); the difference from parsing comes in the existence of a known intended representation of content, the goal tree, against which the emergent parse tree is checked for subsumption at every putative parse step.…”

“…16 Given the perspective on parsing, this cannot be more than an account of tactical generation, associating a tree-structure with a word sequence, rather than strategic generation, the determination of the intended tree-structures from underlying goals. 17 Note that (Otsuka and Purver 2003)'s definition has no explicit context-dependence, for which see (Purver et al 2006). this. Having used computational rules to induce subject and predicate requirements, the word 'John' may be selected to yield a partial tree that subsumes the goal tree, as shown in the parse state in (17).…”

Modelling and Evaluation of Lexical and Syntactic Alignment with a Priming-Based Microplanner

“…According to this model, human language comprehension and production are greatly facilitated by alignment of the interlocutors during conversation. The process of alignment is explained through mutual priming of the interlocutors' linguistic Purver et al [20] take a more formal approach. They use an implementation of the Dynamic Syntax formalism, which uses the same representations and mechanisms for parsing as well as for generation of natural language, and extend it with a model of context.…”

“…The approaches presented above primarily focus on the linguistic and social aspects of alignment in natural language generation. The work of Brockmann et al [9] and Isard et al [13] concentrates on the surface form of language, Bateman [1] sees alignment arising from social-semiotic aspects, and Purver et al [20] are primarily interested in fitting alignment into a formal linguistic framework. In this paper we adopt a more psycholinguistic and cognitive stance on alignment.…”

A: an experimental investigation into...

“…(Pickering and Garrod, 2004, p186) According to this model, speaker and listener ought to be interchangeable at any point, and this is also the stance taken by the grammatical framework of Dynamic Syntax (Cann et al, 2005). In Dynamic Syntax (DS), parsing and production are taken to use exactly the same mechanisms, leading to a prediction that split utterances ought to be strikingly natural (Purver et al, 2006). Additionally, for a third person to process an utterance that appears to come from two separate speakers ought not be more difficult than processing the same utterance from a single speaker, regardless of where in a string the changeover occurs.…”