CAuLD

CAuLD

Construction Automatique de représentations Logiques du Discours

Action de recherche collaborative (ARC) INRIA

Logical Methods for Discourse

Logical Methods for Discourse

December 14-15 2009, Nancy (France)

Presentation

As part of the CAuLD project, this workshop aims at sharing related views on logical and type-theoretical approachs to discourse modelling. This includes aspects related to the underlying formalisms or to the formal modelling of various phenomena.

Participants


Program

(Note that the schedule is still subject to changes.)

Monday Dec. 14

9:00-9:15
 Welcome
9:15-10:00
 A Type-Theoretic Account of Dynamics Philippe de Groote
10:00-10:45
 From DRT to Type Theory and Back
 Jan van Eijck
10:45-11:15
 Break
11:15-12:00
 Semantic Representation of Modal Subordination Using Type Theory(slides : part 1, part 2) Nicholas Asher and Sylvain Pogodalla
12:00-12:45
 Presupposition Projection in Practice Johan Bos
13:00-15:00
 Lunch
15:00-15:45
 Projective Meaning (slides) David Beaver, Craige Roberts, Mandy Simons and Judith Tonhauser
15:45-16:30
 Enriching Contexts for Type-Theoretic Dynamic (slides) Scott Martin and Carl Pollard
16:30-17:00
 Break
17:00-17:45
 Comparative Correlatives and Conditionality
 E. Allyn Smith
17:45-18:30
 Towards a Type-Theoretical Account of Lexical Semantics Christian Retoré

Tuesday Dec 15.

8:45-9:30
 Dynamic Logic in ACG: Discourse Anaphora and Scoping Islands (slides) Oleg Kiselyov
9:30-10:15
 A Dynamic Logic for Presupposition (slides) Ekaterina Lebedeva and Philippe de Groote
10:15-10:45
 Break

10:45-11:30
 Obligatory Presuppositions (slides) Pascal Amsili
11:30-12:15
 Bounded-Rational Theory of Mind for Conversational Implicature (slides) Chung-chieh Shan
12:15-14:00
 Lunch
14:00-14:45
 Vague Updates Joey Frazee and David Beaver

Abstracts

Obligatory Presuppositions

Pascal Amsili

This talk will be about a number of presupposition triggers that seem to be, at least in some cases, obligatory (the most famous example being `too'). We will show that empirically the phenomenon of obligatoriness concerns a not so small subclass of presuppositiontriggers, and try to provide an explanation for this obligatoriness, based on the notion recently put forward of antipresupposition.

Semantic Representation of Modal Subordination Using Type Theory

Nicholas Asher and Sylvain Pogodalla

Following [1]'s proposal of using continuations to model the dynamics of discourse, we propose a new way to build semantic representation of Modal Subordination. We show how different proposals (embedded and not embedded modalities, respective scope of quantifier and modalities, etc.) can be implemented compositionally without any change to the formalism, type theory, using only lexical semantics.

[1]  Philippe de Groote, 2006. "Towards a Montagovian account of dynamics". In proceedings of SALT 16, Tokyo.

Projective Meaning

David Beaver, Craige Roberts, Mandy Simons and Judith Tonhauser

We begin with the observation that so-called "projection tests for presupposition" are not what they appear to be: a wide range of expression types produce inferences that survive embedding, and only a subset of these are inferences that are standardly thought of as presuppositional. One could simply conclude that projection tests are not reliable diagnostics, and move on. Our goal is quite different, to explain why projection occurs, and to provide empirical and theoretical frameworks for studying the full class of inferences that survive embeddings, what we term "Projective Meanings". Our main theoretical claim is that projection is best understood within a theory of information exchange in discourse, and that within such a theory the projective meanings all share one core property, the property of being Not At-Issue. We will outline how this leads to the observed phenomenon of projection, and describe other diagnostic tests which allow us to understand the differences between different types of projective meanings, including differences among standard presupposition triggers.

Presupposition Projection in Practice

Johan Bos

In this talk I discuss theoretical and practical issues of presupposition projection in a wide-coverage text interpretation system. This system implements Discourse Representation Theory on top of Combinatorial Categorial Grammar. It follows Van der Sandt's "presuppostion as anaphora" theory. In particular I will discuss what proper representation of presupposition are, current coverage and inventories of presupposition triggers, and the role of inference and background knowledge.

A Type-Theoretic Account of Dynamics

Philippe de Groote

We provide Montague semantics with a notion of context that allows discourse dynamics to be tackled. The resulting framework subsumes Discourse Representation Theory without appealing to any ad hoc definition. It is based on Church's simply typed lambda-calculus, and the notions of  free and bound variables are as usual. We then define the underlying dynamic logic and show how a Montagovian lexicon can be "dynamized" in a systematic way.

Vague Updates

Joey Frazee and David Beaver

Philosophical analyses of vagueness typically concentrate on the issue of when, or to what extent, vague sentences can be said to be true or appropriate in a given situation. Concentrating on the case of vague scalar predicates, like "tall", we focus instead on the question of what information is conveyed by a vague sentences. Inspired by Williamson's epistemic account of vagueness, we argue for an information theoretic account  which allows for simple logical forms and a completely classical logic, and explains vague effects in terms of statistical uncertainty of various types, notably uncertainty about what degree should be considered tall. We depart from Williamson, and from all previous scholars, in providing concrete measures of the information conveyed by vague sentences. Yet we argue that the account readily applies to standard philosophical concerns, such as the question of when vague sentences are true, why people are reluctant to evaluate borderline cases as definitely true or false, and, of course, issues connected with the sorites.

Dynamic Logic in ACG: Discourse Anaphora and Scoping Islands

Oleg Kiselyov

To analyse scoping islands within the Abstract Categorial Grammar (ACG) formalism we propose an enhancement to ACG along the lines of dynamic logic. The enhanced ACG explains not only the distinct scopes of universals and indefinites, and clause-boundness of universals. We can also apply our ACG to anaphoric indefinite descriptions in discourse. We explain how an indefinite can scope inside negation, yet cannot scope outside negation and create definitedness presuppositions.
Our enhancement to ACG affects only the mapping from abstract language to semantics. We retain all ACG's benefits of parsing from the surface form. Crucially, by avoiding type lifting we keep the order of the abstract signature low, so that parsing remains tractable.
We regard the mapping from abstract language to semantics partial: some sentences, albeit well-formed, just don't make sense. We model this partial mapping as a potentially failing computation in a call-by-value language with multi-prompt delimited control. The evaluation and type inference rules of the language are simple and deterministic. Control prompts may be regarded as loci of binding or quantification, used by quantified phrases and pronouns and set by context. We arrive at the mechanism of interaction of a phrase with its context that determines the scope.

A Dynamic Logic for Presupposition

Ekaterina Lebedeva and Philippe de Groote

Geurt's description theory of names, which is a development of van der Sandt's theory of presupposition, gives elaborated linguistic and philosophical views of proper names semantics. However, it lacks a detailed technical explanation. We attempt to fill this gap by extending de Groote's continuation-based approach of discourse dynamics, which provides lexical semantics within Montague's theory.

Towards a Type-Theoretical Account of Lexical Semantics

Christian Retoré, joint work with Bruno Méry and Christian Bassac

Lexical issues are often left out from logical methods in semantics. We propose here a semantical lexicon which extends and refines the traditional Montague semantics—inspired by work by Pustejovky (the generative lexicon) or Asher (a type driven theory of predication). We remain within a compositional setting expressed by a usual lambda term but other lambda terms whose use is optional enables to solve type mismatch and to access to the various facets of the meaning of a lexical item. Second order lambda calculus is used to anticipate  some type-changes which depend on the type of other lexical items.  We thus provide an algorithm to compute just the correct meanings of compound expressions in the presence of facets or co-predication.
We shall detail how our models yields the correct readings of expressions like:
We shall end with a criticism of this system where the relation between types and terms is too lose while in Asher's approach it might be too tight. Finally we shall shed a light on a possible solution with linear types.

Enriching Contexts for Type-Theoretic Dynamics

Scott Martin and Carl Pollard

Building on work of Heim, Roberts, Muskens, and de Groote, we propose an elaboration of de Groote's notion of LEFT CONTEXT for resolving certain kinds of definiteness presuppositions. Specifically, a left context is a triple consisting of (1) a finite ASSIGNMENT FUNCTION A, the members of whose domain are called DISCOURSE REFERENTS (DR's); (2) a preorder R on the domain of A, called the RESOLUTION; and (3) a (static) proposition C (the COMMON GROUND). In order for a use u of a definite to be felicitous, there must be a unique DR which is greatest (relative to R) among those
DR's which can be inferred from C to satisfy the sortal restriction imposed by u.
This work is part of a joint project with Craige Roberts and E. Allyn Smith on Dynamic Categorial Grammar (DyCG), aimed at the construction of a robust yet practical syntax-semantics-pragmatics interface that deals effectively with projective meaning.

Bounded-Rational Theory of Mind for Conversational Implicature

Chung-chieh Shan, joint work with Oleg Kiselyov

Game-theoretic and relevance-theoretic accounts of pragmatic reasoning often implement Grice's maxim of Manner by penalizing utterances for their syntactic complexity or processing effort.  In ongoing work, we propose to explain this penalty as the _risk of misinterpretation_ in a bounded-rational theory of mind.  The core of this explanation is an executable representation of probabilistic inference in which probabilistic models and inference algorithms are expressed as interacting programs in the same language.  The payoff is that the same cognitive machinery that the hearer uses to decide what to do, by interpreting utterances, can also be used by the speaker to decide what to say, by predicting how potential utterances would be interpreted or misinterpreted by the hearer.

Comparative Correlatives and Conditionality

E. Allyn Smith

English Comparative Correlative (CC) sentences such as The bigger they are, the harder they fall, or The faster we drive, the sooner we'll get there have been compositionally analyzed as conditionals since Beck 1997, predicting that a dynamic theory of conditionals should extend unproblematically to CCs.  This talk evaluates the ways in which CCs are not like conditionals, supplying new data to be captured by dynamic type-theoretical approaches.

From DRT to Type Theory and Back

Jan van Eijck

The talk will sketch a type theoretic version of discourse representation theory and its implementation in Haskell.  Key element in this is the type theoretic treatment of context.  It will be shown how contexts can be enriched with salience information, and how the result can be used to define a more expressive version of DRT.

Reference: Chapter "Discourse Representation and Context" (Chapter 11) in a draft textbook to appear with Cambridge
University Press: "Computational Semantics with Functional Programming."

See: http://www.cwi.nl/~jve/cs/

Practical informations

The workshop will be held at  the LORIA, room B013. There is a map of the campus describing the (pedestrian) way (about 500m)  from the Callot tram stop to the LORIA entrance (see the red arrow).  General informations to access the LORIA are available here. At the reception desk, you should ask for a visitor badge.

General informations on public transportation are available here.

The schedule of the shuttle for the Nancy downtown <-> Lorraine TGV station (XZI) link (about 35 min. You have to buy your ticket in board: 4.50 euros single, 7.40 euros return) is here (note that the schedule will change on Dec 13th. So there might be some small differences. However, synchronisation with TGV will remain). Note that Lorraine TGV station is not the same as Nancy station (which is downtown).

The workshop dinner will take place Monday evening at l'Excelsior, which is on the opposite side of the "place Thiers" from the Nancy station (see the access map).

On Sunday evening, there are very few restaurants that are open. In addition to l'Excelsior, you may also try La Taverne Karlsbrau which is also close to the station, or try your luck around the "place stanislas".

Weather forecast

météo Nancy

Contact