This paper addresses the application of Optimality Theory (OT) in syntax from a computational point of view. The denition of a processing model for OT syntax presupposes a formalization of the notions involved. Here, the work on an OT account based on Lexical Functional Grammar (LFG) can be used as a basis. In order to guarantee that the processing tasks of generation and parsing with an OT grammar are decidable and computationally tractable, certain restrictions have to be imposed on the formalism. The goal of a computational OT approach is to arrive at a sufficiently restricted formalism that is nevertheless expressive enough to capture the intuitions behind the linguistic OT accounts.

Recent theoretical work in LFG proposes a principle-based account of the mapping from cstructure to f-structure. This paper discusses how this account can be applied in a computational context. The assumption of a level of abstraction over rules is required. There are at least two ways of introducing such abstractions without having to alter the architecture of LFG much. Although the generality of the principle-based account is highly desirable not just from the theoretical, but also from an engineering point of view , a naive implementation of the principles will run into complexity problems. However, for most principles, a formulation is possible that avoids these problem without giving up generality of specification.

In the Split NP construction, the descriptive material contributing information about one participant in a predicate's argument structure is realized discontinuously as two nominal phrases in different positions---in the sentence-inital topic position and within the Mittelfeld --, which is a challenge to syntactic accounts of argument selection and unbounded dependency constructions. Most approaches in the literature posit special mechanisms to deal with the construction, and/or fail to explain the parallel behaviour of the Mittelfeld NP and elliptical NPs. I propose an account in Lexical-Functional Grammar making use of standard concepts like unification at the level of functional structure to explain argument doubling (Andrews, 1990), and formalizing semantic composition in terms of linear logic (Dalrymple et al., 1997). The topic NP is explained to be a propertydenoting indefinite (Fanselow, 1988, p. 105), (van Geenhoven, 1996), the Mittelfeld NP is an elliptical NP, anaphorically linked to the topic. Linear logic allows to distinguish the contribution of the two NPs to the sentence meaning although their functional structures are unified, together filling a single argument slot. Through the consequent use of resource accounting in semantic construction, instantiated symbols as values of pred at the functional-syntactic level have been made redundant.

Abstract not found