Agenda


Wednesday: 10.10.
9.45Hans Kamp: Introduction
10.00Yiannis N. Moschovakis: Relative meanings and other (unexpected) applications of the synonymy calculus
By one understanding of Donnellan's views on the referential uses of definite descriptions, there are circumstances when “Smith's murderer is insane” is synonymous with “Jones is insane”—even if Jones is completely innocent. I will give a formal account of this and related phenomena within the theory of referential intensions, and I will use these examples to explain some features of this theory which are not immediately apparent. (The talk will include a brief introduction to referential intension theory and the synonymy calculus.)
11.00Coffee break
11.30Eleni Kalyvianaki: Local meaning and factual content in the theory of referential intensions
In Moschovakis' theory of referential intensions, in addition to global meaning, the local meaning of a sentence at a context of reference is introduced and it is modeled by the abstract algorithm that computes its reference at that context. We present an additional semantical value in this theory, the factual content of a sentence at a context that seeks to capture “what the sentence says about the world at that particular context”. It represents a different algorithm that depends on the way the computation of the reference of each term depends on the context.
12.30–14.00Lunch
14.00Keith Stenning: The plurality of logics necessary for interpreting natural language discourse: an example from the psychology laboratory
The instructions for Wason's famous selection task include the following statements:
“On each card, there is a number on one of its sides and a letter on the other. …
If there is a vowel on one side of a card, then there is an even number on the other side.”
We have proposed elsewhere that Wason implicitly demanded that these two statements be interpreted in different logics—the first defeasible, the second classical—and that this is a substantial source of difficulty for subjects. This talk will review some of the kinds of evidence that this really is a source of difficulty, and then discuss some of the issues that ariese about taking the ‘multiple logics’ stance.
15.00Coffee break
15.30Giosuè Baggio: Aspects of semantic computation in the human brain
Over the past years two related processing principles, immediacy and incrementality, have received increasing support from experimental research in language comprehension. Unfortunately, they appear to be at odds with two important assumptions in formal semantics, namely compositionality and monotonicity (as a property of inference systems). Moreover, immediacy and incrementality seem to resist rigorous formalization, not last because they require an explicit model of language processing at the syntax-semantic interface. In this talk I will present a preliminary sketch of such a model, based on a combination of Performance Grammar (Vosse & Kempen 2000) and the Event Calculus (van Lambalgen & Hamm 2004). Some data on tense processing will be shown, which are consistent with the model and indicate that sentence-level semantic constraints are taken into account as early as 200 ms following the onset of a critical word. I will then turn to one key consequence of immediacy and incrementality, which is non-monotonicity. Some data on aspect processing will be hinted at, suggesting that non-monotonic recomputation is an important on-line aspect of semantic processing. I will conclude bringing all this to bear on the issue of the modularity of meaning computation in the human brain.
16.30Oliver Bott: Processing Consequences of Coercion: the Influence of World Knowledge and Linguistic Context
It is taken for granted that context information plays a crucial role for the interpretation of sentences. However, usually a complete semantic representation of a sentence is derived from syntactic input and context information is integrated after this construction. We will present two experiments which suggest immediate influences of linguistic context and world knowledge on aspectual interpretation. In an ERP study we investigated the reinterpretation of achievements verbs as accomplishments that call for updating purely linguistic information with world knowledge in order to derive intuitively available inferences. In the evoked potentials coercion ellicited a working memory LAN indicating a smooth update of the aspectual representation with an additional preparatory process, but crucially there was no indication of an aspectual mismatch in the ERPs. In a reading experiment we investigated activity sentences within a preceeding dicourse (telic versus atelic). Given a telic context the results indicate that the lexical information of an activity verb can be transformed to an accomplishment thereby completely overriding the Actionsart of the verb. These two experiments suggest that there is a strong interaction between sentence internal and sentence external information during processing, which is incompatible with the two step processing model mentioned above.
Thursday: 11.10.
10.00Luciana Benotti and Patrick Blackburn: Planning and Pragmatics
In this talk we will be presenting recent work on the use of planning in dialogue. As a concrete illustration, we discuss in detail recent work by Luciana Benotti on planning in the setting of a text adventure game; we will view this work as implementing and extending the key concepts of enlightened update (Thomason, Stone and De Vault, 2006). Following this, we will discuss alternative ways of viewing this implementation, namely in terms of accommodations and speech acts.
11.00Coffee break
11.30Johan Bos: Finite Model Building and Computational Semantics
Techniques for finding finite models for first-order theories have been considerably improved in the field of automated deduction in the last years. Nowadays several model builders have been made available by the automated deduction community as off-the-shelf tools, and they have reached a level of sophistication good enough to be of interest for the practical semanticist. Model building has been proposed by several computational semanticists as a tool for disambiguation, preferring interpretations corresponding to minimal models. But the use of model building in semantics goes well beyond this: In practical applications, such as spoken dialogue systems, the nonrecursive structure of finite models form a crystal-clear interface between the abstract semantic representations and the real world; In recognising textual entailment, finite models can support robust inferences.
12.30–14.00Lunch
14.00Peter Koepke and Merlin Carl: The Naproche Project – Linguistics and logic of common mathematical language I.
Mathematical texts are formulated in a semi-formal language, mixing natural language discourse and mathematical formulas. The meaning of a mathematical text can be described by a translation into first-order formulas; the text is correct if its first-order translation is a formal proof in a first-order proof calculus. The Naproche project (Natural language proof checking) recognizes that there is a specific, semi-formal or natural mathematical language which ought to be studied by linguistic techniques. The project aims at constructing a system which checks the correctness of texts written in a controlled but rich sublanguage of ordinary mathematical language including TeX-style typeset formulas. In our talk we demonstrate a small working prototype, explain its modular structure, and discuss future enhancements and extensions, with an emphasis on the mathematical aspects of the system and its applications.
15.00Coffee break
15.30 Bernhard Schroeder und Bernhard Fisseni: The Naproche Project – Linguistics and logic of common mathematical language II.
The special language of mathematics shows some remarkable and specific traits. A controlled language – as used in a proof checker – should cover a broad range of phenomena of the common natural language in the respective area in order to gain high acceptability and usability. A specific feature of the language of mathematics is the use of formal and semi-formal expressions embedded in natural language constructions and the extensive use of variables, which are often syntactically complex. A great part of the text is covered by strongly standardized specific expressions. Among the most characteristic expressions are those introducing assumptions (“Let g be a groupe…”), often associated with the introduction of new discourse entities and new notation. We will discuss an extended DRS format which we employ in our semantic representation of mathematical proofs and which covers these special features of the language of mathematics.
16.30Henk Zeevat: Optimal Minimal Models Optimality Theoretic Pragmatics
Optimality Theoretic Pragmatics gives another and quite natural route to minimal models. But it does more: some properties are maximised that seem hard to capture in purely logical approaches. I will discuss pronouns at some length and discuss the place of plausibility and relevance.
19.00Dinner
Friday: 12.10.
10.00Ulrich Kohlenbach: Unwinding the combinatorial and computational content of proofs
Since the 50's Georg Kreisel proposed to apply proof-theoretic techniques to conrete proofs in mathematics with the aim to extract new information from such proofs by exhibiting the finite combinatorial core of prima facie ineffective proofs. In recent years a systematic approach to this, based on novel forms of proof interpretations, has led to new logical metatheorems as well as numerous applications in nonlinear analysis, fixed point theory, geodesic geometry and ergodic theory. We will present some of the ideas behind these developments.
11.00Coffee break
11.30Peter Schroeder-Heister: Assertion and denial in proof-theoretic semantics, and the square of opposition
Proof-theoretic semantics is an attempt to define logical consequence and, more generally, analytic reasoning in terms of proof rather than truth. By its very nature, in emphasizing proof rather than refutation, it is assertion-driven. It defines what counts as a valid proof of an assertion, and even when it deals with assumptions, it considers them to be placeholders for valid proofs. This is reflected by the fact that in such frameworks negation is defined indirectly by reduction to absurdity rather than by a notion in its own right. Corresponding to the idea of ‘strong negation’ (introduced in 1949 by Nelson), and also corresponding to ideas developed in ‘extended logic programming’, we propose a clausal logic of assertions and denials, where the head of a clause is either an atom or a negated atom. Dealing with reasoning systems of this kind leads to novel symmetry or harmony principles which go beyond the well-known harmony principles for natural deduction or sequent systems. This is due to the fact that, by means of dualization, given (direct) assertion rules lead to associated (indirect) denial rules and vice versa. If we arrange the corresponding four forms of judgment (direct vs. indirect, assertion vs. denial) in the form of a square, their deductive relationships correspond to those exhibited by the traditional square of opposition.
12.30–14.00Lunch
14.00Reinhard Muskens: A Gentzen Calculus for Natural Logic
In this talk I will present an analytic sequent calculus for a fragment of English. Sequents will consist of lambda terms in which all constants are non-logical. Many of such terms closely resemble LF trees for natural language and the system therefore is a form of Natural Logic. The intention is to build an analytic calculus that is reasonably wide-coverage and close to ordinary language, incomplete with respect to the standard interpretation of the terms involved, but strong enough to capture ordinary ‘fast’ reasoning inasfar as it is sound. Many rules in the calculus depend on algebraic properties of certain functor words (such as monotonicity, anti-additivity, antimorphicity, etc., etc.) that also have distributional relevance (Zwarts). I will build on previous accounts of Natural Logic by Van Benthem; Sanchez Valencia; Bernardi; Zamansky, Francez andWinter; and others, but will not base the approach on annotation of proofs in categorial grammar, as is usual in this line of research.
15.00Coffee break
15.30Fritz Hamm and Hans Kamp: The logic of German ung-Nominals
It is well known that German ung-Nominals are ambigous between an object denoting reading, an event denoting reading and a reading which denotes the result of a process. In this paper we will study the ways in which ung-Nominals are dismabiguated by verbal contexts and more generally study the influence of verbal contexts on what inferences can be drawn from sentences containing such nominals. We will propose a combined framework of DRT and Logic Programming in which DRSs are linked via integrity constraints to logic programs. These logic programs provide a non-monotonic inference system for DRT.