Abstract not found

Abstract not found

this paper I shall give a set of simple LFG-like annotated phrase structure rules which connect underspecified representations with other components of the grammar. The rules may be traversed in any order. The underspecified representations will be much like the UDRSs of Reyle [1992], but will formally consist of two parts: (a) a set of equations in ordinary classical type logic with abbreviations that emulate the language of Discourse Representation Theory, and (b) a set of descriptions in the first-order language of trees. Both parts will be generated by the grammar in a relatively independent way. The first part will give us building blocks of `open' Discourse Representation Structures. The second part will summarise all possible ways in which these building blocks may be combined into an ordinary DRS

Categorial Grammars'. Section 4 then continues with a closer look at possible ways to set up a particular Lambda Grammar, lling in some design choices. In particular we will opt for a three dimensional grammar there; one component will deal with dominance and precedence, one with semantics, and one with syntactic features. These choices bring us in close contact with the traditional architecture of Lexical-Functional Grammar (LFG, (Kaplan and Bresnan 1982), for further connections with LFG see (Oehrle 1999) and (Muskens 2001a), which is based upon the present system) and indeed the LFG architecture inspires our answer to question 4 above. Section 4 also works out the logics of the three grammatical components in some detail and thus illustrates one possible set of answers to question 3. For the semantic component we choose a standard type logic with possible worlds; for the feature component a type logic over the rst-order theory of features ((Johnson 1991)); and the multimodal approach to grammar that is found in most modern versions of the Lambek Calculus (see (Moortgat 1997) and references therein) will serve as a basis of the component dealing with dominance and precedence. The multimodal approach is thus moved from the general level of combing signs to one of the special dimensions of the grammar, another illustration of the modularity of the set-up. The chapter ends with a short conclusion.

Abstract. We present a dynamic semantic theory formalized in higher order logic that synthesizes aspects of de Groote’s continuation-based dynamics and Pollard’s hyperintensional semantics. In this theory, we rely on an enriched notion of discourse context inspired by the work of Heim and Roberts. We show how to use this enriched context to improve on de Groote’s treatment of English definite anaphora by modeling it as presupposition fulfillment.

this paper we port a `dynamic'

Abstract. So-called ‘dynamic ’ semantic theories such as Kamp’s discourse representation theory and Heim’s file change semantics account for such phenomena as crosssentential anaphora, donkey anaphora, and the novelty condition on indefinites, but compare unfavorably with Montague semantics in some important respects (clarity and simplicity of mathematical foundations, compositionality, handling of quantification and coordination). Preliminary efforts have been made by Muskens and by de Groote to revise and extend Montague semantics to cover dynamic phenomena. We present a new higher-order theory of discourse semantics which improves on their accounts by incorporating a more articulated notion of context inspired by ideas due to David Lewis and to Craige Roberts. On our account, a context consists of a common ground of mutually accepted propositions together with a set of discourse referents preordered by relative salience. Employing a richer notion of contexts enables us to extend our coverage beyond pronominal anaphora to a wider range of presuppositional phenomena, such as the factivity of certain sententialcomplement verbs, resolution of anaphora associated with arbitrarily complex definite descriptions, presupposition ‘holes ’ such as negation, and the independence condition on the antecedents of conditionals. Formally, our theory is expressed within a higher-order logic with natural number type, separation-style subtyping, and dependent coproducts parameterized by the natural numbers. The system of semantic types builds on proposals due to Thomason and to Pollard in which the type of propositions (static meanings of sentential utterances) is taken as basic and worlds are constructed from propositions (rather than the other way around as in standard Montague semantics).

In standard Montague Semantics we nd a very close correspondence be-tween syntactic and semantic rules (the `Rule-to-Rule Hypothesis'). This is attractive from a processing point of view, as we like to think of syntactic and semantic processing as being done in tandem, with information ow-ing in both directions, from parsing to interpretation and vice versa. The parsing procedure erects the necessary scaolding for interpretation, while semantics (and via semantics context and world knowledge) ideally rules out wrong parses at an early stage. Montague Semantics, however, also favours a strictly bottom up semantic processing architecture. The principle of Compositionality, which says that the meaning of a mother node is to be computed from the meanings of her daughters, seems to enforce such a bottom up procedure. Since we know that parsing algorithms that make use of top down predictions are often much more ecient than those that do not, and since we do not therefore expect human syntactic processing to be strictly bottom up, there is a dilemma. On

Copyright by

an account for sluicing and questions of