A DOP model for Lexical-Functional Grammar

It is well-known that there exist many syntactic and semantic dependencies that are not reflected directly in a surface tree. All modern linguistic theories propose more articulated representations and mechanisms in order to characterize such linguistic phenomena. The Tree-DOP model is thus limited in that it cannot account for these phenomena. In this chapter, we show how a DOP model can be developed for the more articulated representations of Lexical-Functional Grammar (LFG), that is known to be beyond context-free (Kaplan & Bresnan 1982; Kaplan 1989). LFG representations consist of a surface constituent tree enriched with a corresponding functional structure. In order to develop a DOP model for LFG we will define new settings for the four parameters of the general DOP framework, i.e. (1) the representations, (2) the fragments, (3) the composition operation, and (4) the probability model. We will see that the resulting LFG-DOP model triggers a new, corpus-based notion of grammaticality, and an interestingly different class of probability models. We present a parser which uses fragments from LFG