09 Philosophisch-historische Fakultät
Permanent URI for this collectionhttps://elib.uni-stuttgart.de/handle/11682/10
Browse
1 results
Search Results
Item Open Access Economical computation of structural descriptions in natural language(2003) Öhl, Peter Günther; Ian Roberts (Prof., Ph. D.)As indicated in the title, the aim of this paper is to outline a theory enabling us to model an economical way of computing structural descriptions in natural language. This is in fact very much in the spririt of CHOMSKY (1993 etc.) and his followers' Minimalist Program (MP). However, by critically discussing their derivational model and arguing against it, we develop a representational theory that is capable of capturing many relevant regularities of the syntactic and logical structure of sentences by assuming and explaining the interface interactions while interpreting structural descriptions. It is a generative theory of the syntax/semantics interface. We treat all sentences strictly compositionally, starting out from predicates, which we treat as logical functions that are decomposed in syntax, and propositions. Propositions can be marked by logical operators yielding different kinds of modality, also including different clause moods. In the same way we catch the syntax and semantics of subordination. This way of looking at lexical decomposition and markedness through specific operators has very much been inspired by the model of transparent logical form developed by STECHOW (1993 etc.). In order to produce compositional SDs, we use functional application as proposed by MONTAGUE (1969, 1973) for our semantic representation LF. However, we want to avoid the crucial weakness of the kind of categorial grammar he uses (as first proposed by AJDUKIEWICZ 1935) - i.e. the lack of autonomy in the syntactic module. In our model, the syntactic rules do not result from semantic functions. Instead, the representation of semantic formulae is subject to numerous syntactic constraints and principles. Their parameterisation results in syntactic variation, which cannot be explained by categorial grammars. In addition, formal Fs like case, AGR and syntactic categorial Fs play a crucial role for licensing. The crucial role that logical functions play in structure generation is that of semantic licensing. Since in our model all informative items that are inserted into syntax (independent of whether they are represented by words or by opaque elements, i.e. informative 'features' (IFs) that must get transparency indirectly) have lexical entries as logical functions, the composition of a sentence is subject to semantic selection and binding. This might also be expressed in terms of subcategorisation. In order to account for this representationally, we have to allow interaction of the interfaces when a sentence is projected, in order to license the elements inserted into the structure. This is why we call our model a projective licensing grammar (PLG). Our model in fact assumes autonomous modules, but necessarily allows for a little more interaction than GB or the MP. In order to account for structures specific to marked information structure, we also have to allow some interaction with the pragmatic module. As we will show, there is much less controversy in this assumption than syntactic autonomists may think. Another crucial characteristic of PLG is projective economy in the sense of HAIDER (1996a). We want to avoid uneconomical projection of structure, i.e. of phrases that do not host informative elements and are therefore not criterial for the interpretation of the SD. A first step is the reduction of functional phrases by assuming that functional heads are projected only if they represent IFs that express markedness. A second step is allowing functional heads to parametrically host more than one IF. This is reflected in the linear array by the variation of functional complexity of items and word order. In ch. 1 we develop PLG and discuss some of its implications and consequences for syntactic theory. Ch. 2 is dedicated to information structure and the syntax and semantics of subordination. In ch. 3 we apply our theory to the syntax and semantics of clause types. Our analyses are supported by comparative data from 15 languages, each of them representing a number of specific typological properties.