Kolloquium
Pragmatik
Mittwochs,
16:00 - 18:00, C6-200
Gerhard
Jäger
Büro: C6-204
Sprechstunde:
Dienstag 16:15-17:15 oder nach Absprache
&
Hannes
Rieser
Büro: C5-204
Sprechstunde: nach
Vereinbarung
Kursbeschreibung
Das
Kolloquium
befaßt sich mit der Schnittstelle der Pragmatik zu den
anderen
Modulen der Grammatik. Insbesondere sollen die Interaktion zwischen
den Gebrauchsbedingungen sprachlicher Einheiten sowie ihrer Bedeutung
(Semantik) und ihre Struktur (Syntax) beleuchtet werden. Grundlage
für die Diskussion sind in erster Linie die laufenden
Forschungsarbeiten der Teilnehmer selbst, aber auch die aktuelle
Fachliteratur.
Voraussetzung: Einführungsvorlesung
"Semantik und Pragmatik" oder äquivalente Kenntnisse.
Programm
Abstracts
-
Henk
Zeevat (Amsterdam): A Generative Approach to Plural Semantics
Perhaps nowadays we are used
to it, but for the Montague grammarians who in the early '80s read the
first results on the semantics of plural NPs it was a shock: too many
ambiguities, readings that cannot be derived in terms of the normal
decompositions of the sentences and curious invisible operators. It is
fair to say that since then the situation has not improved - the
accuracy of description has, but the conceptual challenge has not been
solved. The talk will attempt to explain plurals from a
constraint-based radical generative semantics perspective. In this
approach, model theoretic objects receive a linguistic realisation that
meets a number of expressive constraints, some of which are determined
by the goals of the speaker. It will be shown that the logical forms
for plural sentences naturally emerge from this approach and that
natural language seem to have developed a fairly reasonable balance
between marking distinctions and leaving them implicit. The talk will
end with some speculation about the incorporation of some observations
about plural anaphora in the framework.
- Christian Ebert (Bielefeld): Expressive
Vollständigkeit und Komplexität unterspezifizierter
Repräsentationen
In den letzten Jahren wurden zahlreiche Ansätze zur
unterspezifierten Bedeutungsrepräsentation skopal mehrdeutiger
Ausdrücke entwickelt. Zwei grundlegende Anforderungen an
solche Formalismen sind (1) die expressive Vollständigkeit,
d.h. die Möglichkeit alle vorkommenden
Skopusambiguitäten auch tatsächlich zu
repräsentieren, und (2) die Vermeidung der kombinatorischen
Explosion, die bei der Berechnung aller Lesarten eines massiv ambigen
Satzes auftreten kann. Erstaunlicherweise wurden trotz ihres
fundamentalen Status beide Anforderungen bislang nicht
ausführlich diskutiert. In meinem Talk werde ich diese
Begriffe diskutieren und drei prominente Ansätze zur
Unterspezifikation evaluieren (Hole Semantics, Normal Dominance
Constraints, Minimal Recursion Semantics), wobei sich zeigen wird, dass
alle drei Formalismen nicht expressiv vollständig sind. Ich
zeige weiterhin, dass es ein prinzipielles Problem mit den obigen
Anforderungen gibt: sie lassen sich nicht beide gleichzeitig
erfüllen. Dieses Ergebnis wirft grundsätzliche Fragen
über die Konzeption und den Nutzen von
Unterspezifikationsformalismen auf.
- Samson
Tikitu de Jager
(Amsterdam):
Monotonicity and the semantics/pragmatics distinction
This talk concerns the two observations (1) that the denotations of
lexicalised generalised quantifiers tend overwhelmingly to be monotonic
(classic examples being "some" and "all"); and (2) that we very
frequently use these quantifiers non-monotonically, via pragmatic
reasoning ("some" interpreted as "some but not all").
Some obvious questions raised by these observations include: why the
pragmatically enriched (non-monotonic) meanings are never lexicalised;
how a child learns the semantic meaning at all, if most normal usage is
pragmatically felicitous; and how the distinction between pragmatic
inappropriateness and semantic falsity is learned, without explicit
signals for discrimination.
I will present a model representing quantifiers as finite state
machines (due to Johan van Benthem [vBen]), and show how incorporating
this in a system of learners (based on the iterated learning model of
Simon Kirby [Kir]) can give rise to the correct predictions regarding
learning.
References
[vBen] J. van Benthem, 1987, "Towards a computational semantics", In P.
Gärdenfors, ed., Generalised quantifiers: Linguistic and
logical approaches, pp. 31--37.
[Kir] K. Smith, S. Kirby & H. Brighton, 2003, "Iterated
learning: a
framework for the emergence of language", Artifical Life
9(4):371--386.
- Stefan Kaufmann (Chicago) Independence in counterfactuals:
premise semantics, causality, and lumping
This talk presents two frameworks for theorizing about counterfactual
conditionals and causality, proposes an integrated analysis, and
concludes with new questions and directions for linguistic research on
counterfactuals.
Ever since Goodman's (1947) groundbreaking work, semanticists have
searched for a general and non-circular way to decide which true
propositions affect the interpretation of conditionals with false
antecedents. The Premise Semantics of Veltman (1976) and Kratzer
(1981) is a widely known formal framework for addressing this
question. The answer remains elusive, however, despite several
attempts over the years to reduce the problem of counterfactuals to
some notion of world-internal structure (Kratzer, 1989, 2003; Kanazawa
et al., 2005; Veltman, 2005).
Meanwhile, a separate line of research in statistics, artificial
intelligence, philosophy and psychology has developed sophisticated
ways of representing causal (in)dependencies, modeling causal
inferences, and testing causal claims empirically (Pearl 2000, among
others). Claims about counterfactuals are frequently made in this
literature, but their relationship to familiar semantic theories is
not always obvious. Moreover, recent psychological research into the
way speakers interpret counterfactuals (e.g., Sloman and Lagnado,
2005) lends only partial support to those claims and seems to call for
substantial refinements.
I will show that an implementation of this sort of causal reasoning in
the framework of Kratzer-style Premise Semantics is both
straightforward and instructive, shedding light on long-standing
empirical disagreements about counterfactuals. In the more
programmatic latter part of the talk, I will point out connections
between recent studies in cognitive psychology and well-established
theories of lexical semantics, which suggest promising avenues for
further psycholinguistic investigation.
Reference
Goodman, N. 1947. The problem of counterfactual conditionals. The
Journal of Philosophy, 44:113-128.
Kanazawa, M., S. Kaufmann, and S. Peters. 2005. On the lumping
semantics of counterfactuals. Journal of Semantics 22:129-151.
Kratzer, A. 1981. Partition and revision: The semantics of
counterfactuals. Journal of Philosophical Logic, 10:201-216.
Kratzer, A. 1989. An investigation of the lumps of thought.
Linguistics and Philosophy, 12:607-653.
Kratzer, A. 2002. Facts: Particulars of information units?
Linguistics and Philosophy, 25:655-670.
Kratzer, A. 2005. Constraining premise sets for
counterfactuals. Journal of Semantics, 22:153-158.
Pearl, J. 2000. Causality: Models, Reasoning, and Inference.
Cambridge University Press.
Sloman, S. and D. Lagnado. Do we 'do'? Cognitive Science 29:5-39.
Veltman, F. 2005. Making counterfactual assumptions. Journal of
Semantics, 22:159-180.
- Alexander Koller (Saarbrücken), Die Skopusunterspezifikation entwächst den Kinderschuhen
Unterspezifikation ist heute der Standardansatz zur Verarbeitung von
Skopusambiguitäten in der Computerlinguistik. In diesem Ansatz, der
seit etwa zehn Jahren entwickelt wird, berechnet man aus einer
syntaktischen Analyse nicht direkt alle semantischen Lesarten, sondern
eine unterspezifizierte Repräsentation (USR), aus der man die Lesarten
bei Bedarf aufzählen kann; dieser Aufzählungsschritt wird aber erst
möglichst spät ausgeführt, um Rechenzeit zu sparen.
In meinem Vortrag werde ich über einige der Fortschritte in der
Unterspezifikation berichten, die sie zu einer ausgereiften
Technologie gemacht haben. So sind heute große Grammatiken mit
unterspezifikationsbasierter Semantikkonstruktion (für MRS) verfügbar,
es gibt effiziente Algorithmen zum Aufzählen von Lesarten (für
Dominanzgraphen), und Übersetzungen zwischen verschiedenen
Unterspezifikationsformalismen (z.B. von MRS in Dominanzgraphen) haben
die formalen Zusammenhänge geklärt und diese Ressourcen jeweils
Nutzern des anderen Formalismus zugänglich gemacht. Diese formalen
Resultate legen auch linguistische Obergrenzen dafür nahe, wie
flexibel die Skopusträger in einer Skopusambiguität vertauscht werden
können. Ich werde weiterhin kurz auf einen neuen Algorithmus eingehen,
der logisch redundante Lesarten aus einer USR eliminiert, ohne sie
auch nur aufzuzählen.
- Ruth Kempson (London), Dialogue Pressures and Syntactic Change
In this talk, I first present a Dynamic Syntax account of word-order
variation in Latin (for both parsing and production), in which I shall
argue that given-new informational effects wherever there is free NP
ordering, together with parallelism effects across conjuncts, are a
consequence of on-line production
pressures in dialogue, explicable in grammar-external pragmatic terms.
I shall go on to show how the effects of these production pressures
become routinised and then encoded in the pronominal clitics of the
medieval Spanish system which through successive steps of routinisation
yield the modern systems with immediately pre-verbal fixed clitic
templates. The arguments turn on being able to explain the
idiosyncratic templatic behaviour of pronominal clitic systems in
terms of their displaying, in calcified form, the range of
structural update mechanisms earlier displayed by NP-construal in general.