Probabilistic and stochastic methods have been fruitfully applied to a wide variety of problems in grammar induction, natural language processing, and cognitive modeling. In this talk I will explore the possibility of developing a class of combinatorial semantic representations for natural languages that compute the semantic value of a (declarative) sentence as a probability value which expresses the likelihood of speakers of the language accepting the sentence as true in a given model. Such an approach to semantic representation treats the pervasive gradience of semantic properties as intrinsic to speakers' linguistic knowledge, rather the result of the interference of performance factors in processing and interpretation. In order for this research program to succeed, it must solve three central problems. First, it needs to formulate a type system that computes the probability value of a sentence from the semantic values of its syntactic constituents. Second, it must incorporate a viable probabilitic logic into the representation of semantic knowledge in order to model meaning entailment. Finally, it must show how the specified class of semantic representations can be efficiently learned from the primary linguistic data available for language acquisition. This research has developed out of recent work with Alex Clark (Royal Holloway, London) on the application of computational learning theory to grammar induction.
Friday September 16, 16.00 - 17.30 Jeanne Peijnenburg (University of Groningen)
Infinite regresses form a problem in many areas of philosophy (metaphysics, ethics, philosophy of mind), but nowhere do they seem to be more pressing than in epistemology. Not for nothing has Laurence Bonjour noted that “considerations with respect to the regress argument [are] perhaps the most crucial in the entire theory of knowledge.” The regress problem in epistemology traditionally takes the form of a one-dimensional epistemic chain, in which (a belief in) a proposition p1 is epistemically justified by (a belief in) p2, which in turn is justified by (a belief in) p3, and so on.
Because the chain does not have a final link from which the justification springs, it seems that there can be no justification for p1 at all. In this talk we will explain that the problem can be solved if we take seriously what is nowadays routinely assumed, namely that epistemic justification is probabilistic
in character. In probabilistic epistemology, turtles can go all the way down.
Friday November 11, 16.00 - 17.30: Reinhard Muskens (Tilburg University) and Noor van Leusen, (Radboud Universiteit Nijmegen)
Events, Times, Worlds, Roles, Linking, and Variable Management
Beaver & Condoravdi 2007 and Eckardt 2010 propose a Linking Semantics for natural language in which it is possible to combine verbs with their arguments in a way that differs fundamentally from the way this is done in Montague Semantics. For example, in Eckardt 2010 lambda abstraction over the verb's arguments is a process that can take place 'on demand' and the possible argument that is abstracted over is determined by the case of the DP that is to be quantified in. The approach has obvious advantages for the treatment of languages that code argument structure with the help of
morphology instead of word order and for the treatment of event semantics, while the need for type raising is reduced.
In this talk, which reports on work in progress, we will point out that some essential ingredients of Linking Semantics are present in type logic once object level axioms for binding are added, as it was done in Muskens 1996. While binding in language may not function in exactly the same way as binding in the logics we are accustomed to, it may still be modeled by these logics if we are prepared to axiomatise it and to thus decouple it from the binding mechanism of the underlying logic. E.g. classical type logic plus some axioms can model the binding in Discourse Representation Theory.
In fact the axiomatisation of binding proposed in Muskens 1996 also enables the logic to talk about related concepts. For example, in Van Leusen and Muskens 2003 it was observed that the notion of a proper DRS (i.e. one having no free discourse referents) can become available once a variant of Zeevat's 1989 semantics is internalised, and Van Leusen 2007 observes that even substitution can be talked about.
It is this last feature that we will depend on here. We will use it to develop a Linking Semantics for Compositional DRT and will discuss strategies to obtain a compositional treatment of language using the rich underlying ontology suggested in our title.
Friday November 25, 16.00 - 17.30: James Woodbridge (Las Vegas)
Revisiting Truth as a Pretense
The central thesis of the pretense account of truth (or rather, truth-talk) is that the notion of truth is a pretense—really there is no such property, we just talk "as if" there were. We make as if to describe things as having or lacking properties called “truth” and “falsity” in order to make claims of other (more complicated) sorts indirectly. The semantic mechanism at work in the operation of truth-talk is a kind of pretense akin to games of make-believe. Despite its involving a kind of fiction, the mechanism of semantic pretense still allows truth-talk to serve as a means for making genuine assertions about the world. Because of how it functions, truth-talk serves to extend the expressive capacity of a language;
centrally, it allows us to make general claims of a special and complicated sort. This has important connections to deflationary accounts of truth. The pretense account of truth-talk also has interesting consequences for other issues, including the liar paradox and the notion of meaning. This talk is a re-examination of the pretense account, in light of the extensions, re-situating, and defenses of the view that have developed over the 6 years since it first appeared in print, through my collaboration with Bradley Armour-Garb.
Monday November 28, 14.00 - 15.30: Wilfrid Hodges (Dartmoor)
Location: Science Park A1.04 (different from the usual location)
How did they manage before they had the notion of quantifier scope?
The expected answer to the title question is: By not mentioning things that require the notion of quantifier scope. By and large this is true. But there are counterexamples, chiefly in the study of sentences with two competing quantifiers, for example Ibn Sina in the 11th century and Peirce in the 19th. Ibn Sina tried to handle these sentences in terms of function dependency, and in this sense his semantics - for natural language sentences - was closer to Henkin and Hintikka than to Tarski.
Friday December 9, 16.00 - 17.30 Nick Asher (Toulouse)
Degrees of Cooperativity
In this talk, I will propose a new way of thinking about conversation from a strategic point of view. I will argue that
a Gricean view simply doesn't do justice to many conversations and that a much more nuanced model of cooperativity is in order. I use game theoretic techniques to develop a model of a minimal or rhetorical kind of cooperativity, which differs
from Gricean content cooperativity.