Abstracts of the Invited speakers

Irene Heim

Not yet available
Department of Linguistics and Philosophy
MIT, Massachusets

Predicate Restriction and Saturation

Bill Ladusaw

At the heart of predicate - argument combination is a calculus of saturation in which the semantically incomplete predicate is made semantically complete. Definite nominal phrases fully saturate the predicate. By contrast indefinite determiner phrases and nominal phrases, which have property contents, can combine with predicates to semantically restrict their denotations without semantically saturating them. In this paper I will discuss cases in which a restricting mode of argument composition seems appropriate.

Many languages present constructions in which the nominal head of an argument is morphologically incorporated into the predicate. Interpreting the incorporated element as restricting the predicate does not eliminate the possibility of further specifying an individual to saturate the predicate. Incorporating languages differ from each other in whether such further specification is allowed. I will argue that the contrast between these languages is due to properties of their syntax and not enforced by the semantics.

In Maori, a Polynesian language spoken in New Zealand, there are two indefinite articles: ``he'' and ``tetahi''. Though they show similar profiles as indefinites and appear to be in free variation in many cases, they each have limitations on their distribution reflects an underlying semantic distinction between them. ``Tetahi'' may not be used in phrases in the pivot of an existential sentence and ``he'' may not occur as the external argument of a verb. I argue that the major features of their distribution follows if we assume that they contrast not in their contents but in the mode in which they compose with predicates. ``Tetahi'' reflects true saturation, with the indefinite representing a choice function. ``He'' combines by predicate restriction. Though the resulting interpretations are generally equivalent, the analysis provides another view of a specific/nonspecific contrast.

This paper is based upon joint work with Sandra Chung.
Department of Linguistics
UCSC, Santa Cruz

The Average American has 2.3 Children

Francis Jeffry Pelletier
(joint work with Greg Carlson and Thomas Hofweber)

Chomsky and certain of his students and followers think it can be demonstrated that the field of semantics, especially the subfield of formal semantics, is bankrupt by considering certain features of some NPs. Hornstein (1984: 58), for example, says: And Chomsky(1995: 29) puts the point and (1986: 45) Although some researchers have tried to counter this sort of argument by postulating some new ³logical forms² for sentences of the relevant sort, most semanticists have reacted to such charges simply by ignoring themŠother than to say that they ³demonstrate a lack of under- standing of formal semantics² on the part of Chomsky and his followers.

In fact, though, I believe there is a real issue here. Firstly, I think that the few attempts to give a direct account of the phenommena do not succeed even on their own terms. And secondly, I think that if the Chomskean charge is not refuted then in fact formal semantics is not a viable enterprise.

I think formal semantics can be saved, but that it will have to give up one of its background assumptions: that the semantic role of NPs is either to denote (designate, etc.) something or to be ³quantificational² in nature. I will try to make this proposal more clear, and will focus on the case of ³average NPs² in doing so.
Department of Philosophy
University of Alberta

Syntactic Constraints on Quantifier Scope Alternation

Mark Steedman

Ambiguities arising from alternations of scope in interpretations for multiply quantified sentences have led to various complications which have compromised the strong assumptions of syntactic/semantic transparency and monotonicity underlying the Frege-Montague approach to the theory of grammar. These include movement at logical form, related abstraction or storage mechanisms, and proliferating type-changing operations. The paper examines some interactions of scope alternation with syntactic phenomena including coordination, binding, and word-order in Germanic languages. Starting from the assumption that many expressions that have been treated as generalized quantifiers are in fact referential expressions, and using combinatory categorial Grammar (CCG) as a grammatical framework, the paper presents an account of quantifier scope ambiguities according to which the available readings follow directly from the combinatorics of the syntactic derivation, without any independent manipulation of logical form and without recourse to type-changing operations other than those with independent syntactic motivation.
Division of Informatics
University of Edinburgh

A Logicist Program for Lexical Semantics

Richmond Thomason

I will begin by trying to clear logicism of the bad name that Frege and Russell gave it. I will situate certain problems in natural language semantics with respect to larger trends in logicism, and will claim that the essence of a good logicist program is a match between a suitable logical formalism and target domain.

I will propose the following logicist program with respect to lexical semantics.

I will illustrate the program with case studies from English derivational morphology. The resulting project is complementary to attempts by Artificial Intelligence logicists to formalize aspects of commonsense reasoning; and it appears to have a philosophical as well as a linguistic dimension.
Department of Philosophy
University of Michigan

Modal Disjunction

Ede Zimmermann

This talk is about a construal of disjunctions "A or B" as lists of epistemic possibilities "It might be that A; it might be that B." Various phenomena can be explained in a natural way if we adopt this non-boolean, modal view of disjunction:
Institut für Deutsche Sprache und Literatur
Johann Wolfgang Goethe-Universität, Frankfurt

Evening Lecture.

Toward a Logic of Perceptions

Lotfi A. Zadeh

Our approach to the logic of perceptions (LP) is inspired by the remarkable human capability to perform a wide variety of physical and mental tasks without any measurements and any computations. Everyday examples of such tasks are parking a car, driving in city traffic, cooking a meal, deciphering sloppy handwriting, summarizing a story, and engaging in discourse.

Underlying this capability is the brain's crucial ability to reason with perceptions -- perceptions of time, distance, force, direction, shape, intent, likelihood and truth, among others. In essence, perceptions are summaries of impressions and as such are intrinsically imprecise.

Perceptions have been studied extensively in a wide variety of contexts. But what is not in existence is a theory in which perceptions are objects of computation. The logic of perceptions (LP) which is outlined in my talk is focused on the development of what is referred to as the computational theory of perceptions (CTP) -- a theory which comprises a conceptual framework for computing and reasoning with perceptions. The base for CTP is the methodology of computing with words (CW). In CW, the objects of computation are words and propositions drawn from a natural language. A typical problem in CW is the following. Assume that a function f, Y=f(X), is described in words as: if X is small then Y is small; if X is medium than Y is large; if X is large then Y is small, where small, medium and large are labels of fuzzy sets. The question is: What are the maximum and maximizing values of Y and X, respectively?

The point of departure in the computational theory of perceptions is the assumption that perceptions are described as propositions in a natural language, e.g., "Michelle is slim," "Mary is telling the truth," "it is likely to rain tomorrow," "economy is improving," "it is very unlikely that there will be a significant increase in the price of oil in the near future." In this perspective, natural languages may be viewed as systems for describing perceptions. In effect, in the computational theory of perceptions computing and reasoning with perceptions is reduced to computing and reasoning with words. Interesting but unrelated approaches to a theory of perceptions have been described by H. Rasiowa and R. Vallee.

To be able to compute with perceptions it is necessary to have a means of representing their meaning in a way that lends itself to computation. Conventional approaches to meaning representation cannot serve this purpose because the intrinsic imprecision of perceptions puts them well beyond the expressive power of predicate logic and related systems. In the computational theory of perceptions, meaning representation is based on what is referred to as constraint-centered semantics of natural languages (CSNL).

A concept which plays a central role in CSNL is that of a generalized constraint. Conventional constraints are crisp and are expressed as X is C, where X is a variable and C is a crisp set. In a generic form, a generalized unconditional constraint is expressed as X isr R, where X is the constrained variable; R is the constraining (fuzzy) relation which is called the generalized value of X; and isr, pronounces as ezar, is a variable copula in which the value of the discrete variable r defines the way in which R constrains X. Among the basic types of constraints are the following: equality constraints (r:=); possibilistic constraints (r:blank); veristic constraints (r:v); probabilistic constraints (r:p); random set constraints (r:rs); usuality constraints (r:u); fuzzy graph constraints (r:fg); and Pawlak set constraints (r:ps).

In constraint-centered semantics, a proposition, p, is viewed as an answer to a question, q, which is implicit in p. The meanings of p and q are represented as generalized constraints, which play the roles of canonical forms of p and q, CF(p) and CF(q), respectively. CF(q) is expressed as: X isr ?R, read as "What is the generalized value of X?" Correspondingly, CF(p) is expressed as: X isr R, read as "The generalized value of X isr R." The process of expressing p and q in their canonical forms plays a central role in constraint-centered semantics and is referred to as explicitation. Explicitation may be viewed as translation of p and q into expressions in GCL -- the Generalized Constraint Language.

In the logic of perceptions, representation of meaning is a preliminary to reasoning with perceptions -- a process which starts with a collection of perceptions which constitute the initial data set (IDS) and terminates in a proposition or a collection of propositions which play the role of an answer to a query, that is, the terminal data set (TDS). Canonical forms of propositions in IDS constitute the initial constraint set (ICS). The key part of the reasoning process is goal-directed propagation of generalized constraints from ICS to a terminal constraint set (TCS) which plays the role of the canonical form of TDS. The rules governing generalized constraint propagation in the logic of perceptions coincide with the roles of inference in fuzzy logic. The principal generic rules are: conjunctive rule; disjunctive rule; projective rule; surjective rule; inversive rule; compositional rule; and the extension principle. The generic rules are specialized by assigning specific values to the copula variable, r, in X isr R.

The principal aim of the logic of perceptions is the development of an automated capability to reason with perception-based information. Existing theories do not have this capability and rely instead on conversion of perceptions into measurements -- a process which in many cases is infeasible, unrealistic or counterproductive. In this perspective, addition of the machinery of the computational theory of perceptions to existing theories may eventually lead to theories which have a superior capability to function in an environment of imprecision, uncertainty and partial truth.
Department of EECS
University of California

Abstracts Accepted for Presentation at the Twelfth Amsterdam Colloquium.

Focus and topic sensitive operators

Maria Aloni, David Beaver and Brady Zack Clark

This paper concerns so-called focus sensitive particles (FSPs), which appear to encode meanings sensitive to intonation. It will be argued that of the wide array of particles previously described as focus sensitive fall into two natural subclasses cross-linguistically. We use a range of original data, especially extraction and presupposition phenomena, to show that only one of the resulting classes is genuinely focus sensitive. It will further be argued that no current theory of focus can adequately account for both sub-classes.
ILLC/Department of Philosophy
University of Amsterdam
and
Department of Linguistics
Stanford University

Two place probabilities, full belief and belief revision

Horacio Arlo-Costa and Rohit Parikh

Making use of van Fraassen's framework for defining beliefs from conditional probabilities, we provide a discussion of the issues and present some new technical results. We provide a complete characterization of van Fraassen's probabilities for countable spaces, show how belief revision can be defined naturally, even by propositions of probability 0, and give some axioms which are sound for the framework.
Computer Science, Mathematics and Philosophy
City University of New York
and
Philosophy Department
Carnegie Mellon University

The logic of anaphora resolution

David Beaver

This paper concerns the semantics/pragmatics interface for natural language, and in particular the question of how anaphora resolution should be orchestrated in a dynamic semantics. Previous dynamic systems such as DPL have relied on preindexation of anaphors and antecedents. It is argued that this represents a serious inadequacy. A dyanmic semantic system, RPL, is proposed which eliminates the need for preindexation, by combining an "Amsterdam-style" dynamic semantics with a pragmatic module based on the Centering algorithm. The semantics uses a novel extension of DMPL information states, in which ambiguity of anaphors is represented using multiple referent systems. The pragmatic componenent also uses a novel approach, reformulating Centering declarations as a preference order over dynamic transitions. It is argued that the resulting system not only provides a marked empirical imporvement over dynamic predecessors, but also provides a quite general approach to the semantics/pragmatics interface.
Department of Linguistics
Stanford University

Plural predication and partitional discourses

Sigrid Beck

This paper argues against Schwarzschild's (1996) proposal that the interpretation of relational plural sentences is constrained by a contextually determined relation. Instead, I suggest that the data that seems to indicate the use of such relations are to be reanalysed as a discourse phenomenon related to telescoping, which I term partitional discourses. This captures the fact that the only relations that seem to be available are equivalence relations. If there is no restricting salient relation, Schwarzschild's uniform analysis of relational plurals in terms of double universal quantification becomes untenable. We come back to Krifka's (1989) proposal of a polyadic plural operator to handle cumulative readings of relational plurals. This theory of plural predication has to be embedded in a semantic framework capable of ccounting for discourse effects.
Department of Linguistics
University of Connecticut

Incorporation as unification

Agnes Bende-Farkas

This paper proposes an analysis of a particular Definitesness Effect construction, that involving the light verb {\em have} as found in e.g.\ {\em John has a sister}, or {\em mary has a good salary}. The analysis is based on the notion of term unification, which has played an important role in computer science and computational linguistics but, to my knowledge, hardly within formal semantics. According to term unification, the verb {\em have} and its object phrase boh introduce a predicational ``term'' consisting of a predicate and its arguments. Combining verb and object phrase involves unifying these two terms. When, for some reason or another, unification fails, the combination is unacceptable. The analysis seems capable of extension to other Definitesness Effect constructions, including those involving ``light'' verbs in Hungarian.
IMS
Universit\"at Stuttgart

Questions as first class citizens

Martin van den Berg

In this paper we use the Linguistic Discourse Model and Dynamic Quantifier Logic (van den Berg 1996, van den Berg and Polanyi 1999) to give a formal treatment of questions and answers in dynamic semantics. Following the ideas developed in van den Berg 1996, we give a treatment of wh-phrases as a form of generalized dynamic quantifier that explains how questions can contain anaphoric references and introduce new ones. Our approach, which stays close to the account of questions and answers in (Groenendijk and Stokhof 1984), implements incremental interpretation following the Linguistic Discourse Model.

FX Palo Alto Laboratory

A new probability model for data oriented parsing

Remko Bonnema, Paul Buying and Remko Scha

{\em Date oriented parsing systems} employ redundant Stochastic Tree Substitution Grammars to analyse natural language utterances on the basis of an annotated corpus (a {\em tree-bank}). A fundamental component of such systems is the way in which the substitution-probability of a tree is estimated from its ocurrences in the treebank. In the standard method for doing this, the probability of a tree is directly correlated with its occurrence frequency in the bag of all fragments of all corpus trees. We show that this results in undesirable statistical biases. We therefore propose an alternative method, which estimates the substitution-probability of a fragment as the probability that it has been involved in the derivation of a corpus tree. We show that this method has more plausible properties.
Institute for Logic, Language and Computation
University of Amsterdam

Consequences from Quine

Robin Clark and Natasha Kurtonina

We reconstruct Quine's (1960) combinatory logic (QCL), providing both a semantics and a proof theory for it. We, furthermore, derive the surprising result that QCL can be used as a framework for dynamic semantics, in particular allowing for a compositional account of reference tracking mechanisms in natural language, including switch reference. Finally, we demonstrate that QCL can express the dynamic modalities of arrow logic.
Department of Linguistics
University of Pennsylvania
and Institute for Research in Cognitive Science
University of Pennsylvania

Reciprocal interpretation with functional pronouns

Alexis Dimitriadis

Reciprocal constructions in which the antecedent of the reciprocal is a dependent pronoun have been analyzed as instance of ``wide-scope'' reciprocals. Because the binder of the reciprocal determines its range, this analysis cannot handle sentences in which the antecedent of the reciprocal is not bound by a co-refering, c-commanding antecedent. I propose to derive the correct range for the reciprocal by translating pronouns as functions (in the fashion of Jacobson (1999)). The range of the reciprocal antecedent is computable by applying a maximality operator to the restricted function representing its antecedent.
Department of Linguistics
University of Pennsylvania

The semantics of transitivity alternations

Edit Doron

In theoretical linguistics, causative and middle verbs are usually derived by independent operations. But cross linguistically, both mark the same transitivity alterations. This paper proposes a unified syntactic system for the derivation of both types of verbs, which, moreover, sheds new light on problems in the interface of semantics and morphology. One problem is the impossibility, mostly ignored in linguistic theory, of deriving the semantics of middle verbs from that of the corresponding transitive verbs. The second is explaining the identity found cross linguistically between middle and reflexive morphology. The third is providing an alternative to the ``event-decomposition'' account of causative verbs.
Department of English Languages
Hebrew University of Jerusalem

Deriving and resolving ambiguities in {\em wieder\/}-sentences

Markus Egg

The paper discusses two challenges posed by sentences with German {\em wieder} `again', viz., the {\em derivation} and the {\em resolution} of the `repetitive'/`restitutive' ambiguity that is characteristic for them. It is shown that the framework of underspecification is adequate for both these tasks. Underspecification allows for an easy derivation of these ambiguities that takes into account syntactic restrictions for semantic ambiguities. The resulting semantic representations are suitable input for resolution processes that resolve the ambiguity of {\em wieder}-sentences.
Computerlinguistik, Geb. 17.2
Universit\"at des Saarlandes

Non-monotonicity from constructing semantic representations

Tim Fernando

A standard approach to non-monotonicity locates the phenomenon in preferences between models, against which certain well-formed formulas (or semantic representations, SRs) are interpreted. Such an approach skips over the prior step of constructing suitable SRs for natural language discourse --- arguably the main challenge from the perspective of formal linguistics (computational or otherwise). The present work focusses on this step, tracing complications of presupposition and ambiguity to it. A family of modal logics is outlined, supporting an analysis of non-monotonicity as the failure of a sentence-by-sentence translation of a sequence of natural language sentences to persist; that is, the SR associated with a sentence may, in light of further natural language input, need to be revised. This revision may involve adjustments to background assumptions, implicated in presupposition accommodation.
Computer Science
Trinity College Dublin

The epistemics of presupposition

Anthony S. Gillies

Successful linguistic interaction requires agents to be able to reason about what is being said. Given this, it is natural to pursue the extent to which general epistemic principles can be brought to bear on semantic phenomena. In particular, I look at the interaction between the dynamics of belief and presuppositon. A simple defeasible update semantics for presupposition is given in which accommodation is understood as belief updating, and presupposition failure as failed belief revision.
Department of Philosophy
University of Arizona

Constructional ambiguity in conversation

Jonathan Ginzburg and Ivan Sag

The paper considers the contents associated with clauses in conversational interaction, focussing particularly on the issue of how grammatical frameworks can accommodate uses such as intonation questions and reprise uses. We point out that the phenomena at issue are problematic for views of the syntax/semantics interface such as radical lexicalism and syntactic modularity. We decribe a grammatical system formulated within HPSG in which generalizations about phrasal types, characterized by means of constraints on both the syntactic and semantic components of the sign, are captured by means of multiple inheritance hierarchies. We show how this system can accommodate the relevant phenomena.
Department of Linguistics
Stanford University
and
Dept of Philosophy, King's College, London and
Dept of English, Hebrew University of Jerusalem

Cross-linguistic semantics of weak pronouns in doubling structures

Javier Guti\'errez-Rexach

In this paper, a formal semantics of prosodically weak pronouns or clitics is developed in which they are treated as a type of generalized quantifier inherently restrictied to a context set (following Westerst\aa hl (1985, 1989) and van der Does (1995) general treatment of strong pronouns). It is claimed that cross-linguistic variation in ``clitic doubling'' configurations, where the pronoun has a quantifier associate, can be accounted for in the terms of GQ theory and emerges from the existence of a series of semantic parameters on the retrievAl of context sets.
Cunz Hall
Ohio State university

Substructural logic: a unifying framework for second generation datamining algorithms

Erik de Haas and Pieter Adriaans

In this paper we propose a framework for data mining algorithms based on a system of substructural logic. We show the connections between data mining, inductive logic programming and grammar induction. Furthermore we present a small family of substructural logics that can represent modern complex information systems. This small family of substructural logics on its turn will enable us to design efficient datamining algorithms using techniques from the field of inductive logic programmaing.
ILLC/WINS
University of Amsterdam and Syllogic

Toward a unified analysis of DP conjunction

Caroline Heycock and Roberto Zamparelli

In this paper we analyse ``split'' conjunction within DPs, as found in English examples such as (a) {\em that man and woman} and (b) {\em those cats and dogs}. We demonstrate that many of the well-studied Western European languages allow plural split conjunction (b) but not singular (a); in these languages DP-internal singular conjunction can only have the intersective ``joint'' reading possible also in English examples like (c) {\em my friend and colleague}. We show that the operation of ``set product'' (union over each possible $n$-tuple across $n$ conjuncts) can be used to define conjunction in a way that derives all the available readings. In order to explain their cross-linguistic distribution we propose a minimal difference between languages in the way they obtain the denotation of a singular noun phrase. Taken together, these ideas yield a syntactic and semantic theory of conjunction which can not only account for the DP data but also explain the distributivity properties of different conjoined categories.
Department of Theoretical and Applied Linguistics and
Human Communication Research Centre
University of Edinburgh

Deconstructing Jacobson's {\bf Z}

Gerhard J\"ager

Following the methodology of Moortgat 1996b, we compare two multimodal deconstructions of Jacobson's 1999 type shifting operator {\bf Z} for anaphoric dependencies. The global analysis of J\"ager 1998 is shown to be of limited generality: it restricts the occurrence of anaphoric expressions to {\em associative} environments---environments where sensitivity for constituent structure is lacking. We propose an alternative deconstruction where anaphora resolution is independent of resource mangement assumptions about the structural context. The analysis is based on the general theory of structural control proposed in Kurtonina and Moortgat 1995: the interaction principles of two binary products are fine-tuned in terms of unary modal controol devices.

Zentrum fuer Allgemeine Sprachwissenschaft

IF logic and informational independence

Theo M.V. Janssen

In game theoretical smeantics the truth of a formula is determined by a game between two players, $\forall $belard who tries to verify the formula, and $\exists $loise to refute it. He chooses on $\wedge $ and $\forall $, she on $\vee $ and $\exists $. A version of such games, introduced by J.\ Hintikka, is IF logic: independence friendly logic. The quantifier $\exists (y/x)$ means that $y$ has to be chosen independent of $x$, and $\psi (\vee /x)\theta $ that a subformula has to be chosen independent of $x$. A formula is true, if $\exists $loise has a winning strategy. Hodges has given a compositional interpretation for the logic: trump semantics. It will be argued that this interpretation gives results that are not in accordance with intuitions concerning indepedence of information. Two -equivalent- alternative interpretations will be proposed that do correspond with intuitions, one based on playing games, and one on sets of assignments.
Computer Science
University of Amsterdam

A calculus for direct deduction with dominance constraints

Jan Jaspars and Alexander Koller

Underspecification has recently been a popular approach to dealing with ambuity. An important operation in this context is {\em direct deduction}, deduction on underspecified descriptions which is justified by the meaning of the described formulae. Here we instantiate an abstract approach to direct deduction to dominance constraints, a concrete underspecification formalism, and obtain a sound and comple calculus for this formalism.
Department of Mathematics and Computer Science
University of Amsterdam
and
Department of Computational Linguistics
Universit\"at des Saarlandes

True to Facts

Jacques Jayez and Dani\`ele Godard

Ever since Vendler, the denotation of the term {\em fact} has been considered problematic. Facts seem to be entities of the world and informational items at the same time, located somewhere between states of affairs or events and propositions. Using Zalta's theory of abstract objects, we investigate where this somewhere is. We propose that facts are abstract objects which {\em encode} (in Zalta's sense) the property of being such that a given proposition holds. Thus, the correspondence between facts and propositions is not an external relation between two types of entities, but a built-in property of facts. This allows us to distinguish facts from propositions and events but also to connect those categories and to acount for their linguistic interplay.
EHESS and
CNRS and Lille III

Semantic composition for partial proof trees

Aravind K. Joshi, Seth Kulick and Natasha Kurtonina

We address the problem of semantic composition in a categorial system based on a hybrid logic. One logic is used for unfolding a categorial type, resulting in a partial derivation. The second logic computes the semantic representation from those partial derivations. We encode the history of the derivation from the first logic by using bound variables to represent the missing assumptions. Since the application of the partial derivations can take place in either direction in the second logic, this results in $\lambda $-terms inside the resulting semantic term. We show that by allowing such terms to be moved to the outermost position, then compositionality can be maintained in the hybrid logic.
Institute for Research in Cognitive Science
University of Pennsylvania

A proof-theoretic view of intensionality

Reinhard Kahle

We discuss a proof-theoretic view of intensionality. Based on a notion of the {\em use} of a formula in a proof we show how a proof-theoretic account can avoid some well-known difficulties of the representation of intensional phenomena. The key example is binary necessity, where we read {\em ``A is necessary for B.''} as {\em ``Every proof of B uses A.''} Provided that {\em A} is an axiom or an atomic proposition we can give a formalized version of this reading. This theory is compared with the standard modal logic approach to necessity and two examples are given. Finally we give an outlook over further applications of the proof-theoretic view of intensionality which turn out to be a nice example of interdisciplinarity between logic, philosophy, linguistics and computer science.
WSI
Universit\"at T\"ubingen

Factoring predicate argument and scope semantics: underspecified semantics with LTAG

Laura Kallmeyer and Aravind Joshi

This paper proposes a compositional semantics for lexicalized tree-adjoining grammar (LTAG). The use of tree-local multicomponent derivations allows separation of the semantic contribution of a lexical item into two parts: one component contributes to the predicate argument structure whereas the second component contributes to scope semantics. Starting from this idea a syntax-semantics interface is presented where the composition of a semantic representation depends only on the derivation structure. It is shown that the derivation structure (and indirectly the restrictions resulting from the locality of the formalism) allows an appropriate amount of underspecification. This is illustrated by showing the generation of underspecified semantic representations for quantifier scope ambiguities.
Sonderforschungsbereich 441 and IRCS
Universit\"at T\"ubingen and University of Pennsylvania
and
Institute for Research in Cognitive Science
University of Pennsylvania

The dynamics of tree growth and quantifier construal

Ruth Kempson and Wilfried Meyer-Viol

This paper demonstrates how quantifier-construal can be globally defined without sacrificing an incremental left-right process of projecting interpretation. Scope statements, collected during the course of the left-right parsing process and subject to lexically specified restrictions, are separated from the projection of content for quantifying expressions. Amongst other scope restrictions, construal of indefinites is shown to resemble pronominal anaphora, with indefinites defined as taking narrow scope with respect to some element to be chosen, a free pragmatic choice determining the scope relation feeding algorithmic processes determining the content of quantified expressions.
Philosophy Department
King's College London

Relating polyadic quantifiers: on the cumulativity-distributivity interplay

Brenda Kennelly and Fabien Reniers

We show how to relate tri-adic quantifiers that express mixed readings of distributivity and cumulativity within a single 3-place predicate to the dyadic quantifiers that express distributivity (function composition of monadic quantifiers) and cumulativity (Scha 1981). We discuss problems with the standard approaches and propose that cumulativity necessarily takes precedence over distributivity. Consequently, for mixed readings cumulativity is reanalyzed as a relation between a type $<2>$ and a type $<1>$ quantifier. This new account of cumulativity generalizes conveniently to cumulative quantifiers of arbitrary type.
OTS
University of Utrecht

Using centering theory to plan coherent texts

Rodger Kibble and Richard Power

Centering theory (CT) has been mostly discussed from the point of view of interpretation rather than generation, and research has tended to concentrate on problems of anaphora resolution. This paper examines how centering could fit into the generation task, concentrating on the implications for {\em planning} of texts and sentences. We show that the CT rules as they stand do not fit neatly into a ``consensus'' pipe-lined NLG architecture as they appear to rely on feedback from surface grammatical relations to text planning. We suggest some ways this problem can be overcome or circumvented and report on initial implementation efforts.
ITRI
University of Brighton

Identification language games

Peter Krause

Identification dialogues are inquiries into which individual a speaker intends to refer to. Typical identification dialogues occur when a riddle is solved and when a reference is clarified in a subdialogue. Here, the simplest kind of identification dialogue is studied abstractly. A version of discourse representation theory (DRT) with presuppositions and an epistemic operator; and a specification of the propositional attitudes of the participants together with anchoring relations are used to specify the scoreboard of a language game of identification dialogues with two participants, the identifier and the informant. The rules are based on the standard semantics for the representations. A {\em defeasible} notion of {\em knowing which} object is meant can be formulated. The meaning of the locution {\em which one X is depends on which one Y is.} is specified.
Institute for Computational Linguistics
University of Stuttgart

Binding by implicit arguments

Alice ter Meulen

A game-theoretic account of anaphoric definite descriptions is first presented for discourse binding with an implicit existential argument. Verifier¹s claimed existence of a verifying strategy for the inferred existential statement explains why pronouns cannot be bound by implicit arguments, until the Falsifier demands execution of that claimed verifying strategy. It is discussed how DRT, as representational modeltheoretic dynamic semantics, and DPL, as compositional modeltheoretic dynamic semantics, would differentiate between asserted, inferred and presupposed indefinites to account for the observations. Implicit arguments are linguistically economical as they circumvent scope-disambiguation, forcing the inferred existential to remain in focus, and constrain accommodation of presuppositions.
Department of English
University of Groningen

Modeling ambiguity in a multi-agent system

Christof Monz

This paper investigates the formal pragmatics of ambiguous expressions by modeling ambiguity in a multi-agent system. Such a framework allows us to give a more refined notion of the kind of information that is conveyed by ambiguous expressions. We analyze how ambiguity affects the knowledge of the dialogue participants, and, especially, what they know about each other after an ambiguous sentence has been uttered. The agents communicate with each other by means of a {\tt tell}-function, whose application is constrained by an implementation of some of Grice's maxims. The multi-agent system itself is represented as a Kripke structure and {\tt tell} is an update function on those structures. This framework enables us to distinguish between the information conveyed by an ambiguous sentence vs.\ the information conveyed by disjunctions, and between semantic ambiguity vs.\ perceived ambiguity.
Institute for Logic, Language and Computation
University of Amsterdam

DPL with control elements

Rick W.F. Nouwen

Focus in conditionals can cause a deviating external dynamic behaviour. To account for these exceptional cases, we constructed a variant of dynamic predicate logic, which uses control elements to direct information in need of a special treatment to a special location in information states. On the top-level meaning consists of various relations each contributing some relevant part of the semantics. Further down meaning is still given by means of ordinary DPL-relations. The result is a fully incremental logic, which can be used to construct meanings straight from the natural word order. The system shows that the construction of variants of DPL can be very useful.
UiL-OTS
University of Utrecht

Semantics for attribute-value theories

Rainer Osswald

First, we show how to reconstruct attribute-value (AV) descriptions from natural language by regimentation and formalization within first-order predicate logic. The introduction of appropriate predicate operdators then leads to AV-expressions of the usual Kasper-Rounds type. We present a slight extension which permits relations between attribute values. A straightforward modification of standard AV-logic turns out to be sound and complete with respect to first-oprder derivability granted that attributes are functional. Demonstrating this is part of our second concern which is to apply geometric logic and locale theory to AV-theories like HPSG. Viewing AV-theories as propositional geometric theories provides a crisp characterization of the denotation of an AV-theory as the point space of its classifying locale.
Informatikzentrum
Fernuniversit\"at Hagen

Modeling coalitional power in modal logic

Marc Pauly

We introduce a modal logic for describing what groups of agents can bring about. Using a neighborhood semantics, we arrive at the notion of a {\em coalitional model} which associates with every group of agents the sets of states for which the coalition is effective. We show how specific classes of coalition models or frames can be used to obtain a dynamic model of collective action in (1) strategic game forms and (2) extensive game forms of perfect information. These results draw upon the study of {\em effectivity functions} in social choice theory and generalize some of the results obtained here.

Centrum voor Wiskunde en Informatica

Questioning to resolve decision problems

Robert van Rooy

Why do we ask questions? Because we want to have some information. But why this particular kind of information? Because only information of this particular kind is helpful to resolve the {\em decision problem} that an agent faces. In this paper I argue that questions are asked because their answers help to resolve the questioners decision problem. By relating questions to decision problems I show (i) how we can measure the values of questions, and (ii) how answers can resolve questions in particular circumstances, although the nswer is not exhaustive and complete in the sense of Groenendijk and Stokhof (1984).
Department of Philosophy
University of Amsterdam

Towards a semantic-based theory of language learning

Isabelle Tellier

The notion of Structural Example has recently emerged in the domain of grammatical inference. It allows to solve the old difficult problem of learning a grammar from positive examples but seems to be a very {\em ad hoc} structure for this purpose. In this artcile, we first propose a formal version of the Principle of Compositionality based on Structural Examples. We then explain under which conditions the Structural Examples used in the domain of grammatical inference can be obtained very easily from sentences and their semantic representations, which are naturally available in the environment of children learning their mother tongue. Structural Examples thus appear as an interesting intermediate representation between syntax and semantics. This leads to a new formal model of language learning where semantic information play a crucial role.
Laboratoire d'Informatique Fondamentale de Lille
Universit\'e Charles de Gaulle-Lille3

Two approaches to modal interaction in discourse

C.F.M. Vermeulen

We have seen several examples of the interaction of modal expressions in discourse. The intuitive explanation of this interaction is that a modal depends on a contextually given set of possibilities {\em and} adds a new set to the context, that can be used by subsequent modalities. \par Two style of formalisation of this explanation were considered, both in dynamic semantics. One uses a representation language with indexed modalities and gives an update semanticsfor this language. The other, `algebraic' approach uses a string language with an interpretation in an algebra of m-{\sc states}. Both styles allow us to represent the crucial example. For a more general comparison we have a systematic translation between the formalisms that preserves meaning. We conclude that the algebraic approach is at least as well suited for the representation of modal interaction in discourse. In addition the algebraic approach forces us to avoid magical updates and thus ensures a level of computtional realism that update semantics {\em per se} does not offer.
Departments of Philosophy and Linguistics
Utrecht University

A different game? Game theoretical semantics as a new paradigm

Louise Vigeant

Hintikka claims that Game Theoretical Semantics (GTS) is different enough from the dominant language theories, e.g. Montague Semantics, to represent a new paradigm in the Kuhnian sense. GTS differs in the three following ways: minimal syntax, non-compositional and a definition of truth in terms of strategy. Of these three only the strategic definition of truth proves to be an irreducible difference between the two types of semantic theories. This new definition of truth is not enough to define GTS as a different paradigm. What leads Hintikka to claim that GTS is a new paradigm is his belief that this semantic theory is representative of Wittgenstein¹s language games. One of the main arguments in his interpretation of Wittgenstein is that language games are a major innovation in explaining rules. The attempt to incorporate this innovation into a semantic theory is only a partial step towards a new type of semantic theory. A Wittgensteinian paradigm must also answer the question of what it is about truth that Œfits¹ a proposition.
Department of Linguistics
Stanford University

Plural type quantification

Yoad Winter

This paper develops a type theoretical semantics for quantification with plural noun phrases. This theory, unlike previous ones, sticks to the standard treatment of singular quantification and uses only one lifting operator per semantic category (predicate, quantifier and determiner) for plural quantification. Following Bennett (1974), plural individuals are treated as functions of type $et$. Plural nouns and other plural predicates accordingly denote $(et)t$ functions. Such predicates do not match the standard $(et)((et)t)$ type of determiners. Following Partee and Rooth (1983), type mismatches are resolved using {\em type shifting operators}. These operators derive collectivity with plurals, keeping the analysis of singular noun phrases, where no type mismatch arises, as in Barwise and Cooper (1981). A single type shifting operator for determiners combines into one reading the {\em existential} shift and the {\em counting} (neutral) shift of Scha (1981) and Van der Does (1993). This operator combines the {\em conservativity} principle of generalized quantifier theory with Szabolcsi's (1997) existential quantification over {\em witness sets}. The unified lift prevents unmotivated ambiguity as well as the monotonicity ill of existential lifts pointed out by Van Benthem (1986:52-53).
Computer Science
Technion

A dynamic solution for the problem of validity of practical practical inference

Berislav \v {Z}arni\'c

The paper develops a dynamic framework suitable for analysis of practical propositional inference. The application of functional approach in dynamic semantics confirms the results obtained in philosophical analyses of practical inference, regarding, in particular, defeasibility, undetachability and distinctive quality of conclusion. Contradictory results of proposed tests of validity and conflicting intuitions on valid forms of practical inference are reconciled.
Teacher's College
University of Split

Explaining presupposition triggers

Henk Zeevat

The paper proposes three revisions to the standard view of presuppositions: the employment of optimality theory for the defaults and preferences, the admission of weak antecedents for presupposition resolution/satisfaction and a fine-grained classification of presupposition triggers, based on the availability of expression alternatives and the logical requirement of the presupposition. The treatment is able of dealing with a wide range of phenomena that are outside the scope of any current presupposition theory.
ILLC/Computational Linguistics
University of Amsterdam


Paul J.E. Dekker