In this paper we pursue two related goals: First, we establish a conceptual link between chunk-based syntactic structures as typically assumed in shallow parsing approaches, as opposed to principle-based syntactic structures as assumed in theoretical linguistics research. This conceptual link emerges from the study of configurational vs. non-configurational languages, their analysis within the LFG framework, and the observation of diverse strategies for ambiguity resolution across this spectrum of (non-)configurational language types. Second, we show how shallow analyses as usually employed in practical NLP applications can be refined to deliver full-fledged syntactic representations, by designing an architecture for LFG f-structure projection from chunk-based syntactic analyses.
In line with our two-fold goal we will demonstrate that principles for f-structure projection from chunks are similar -- modulo specific attachment constraints -- to the LFG analysis of non-configurational languages. In essence, then, besides the design of a new style of robust LFG processing from chunk-based analyses, our investigation offers theoretical insight into the kind of abstraction (i.e. underspecification) employed in shallow analysis, and how it can be formalised within the LFG framework. In particular, we will show how to adapt the LFG analysis of non-configurational case-stacking languages in terms of inside-out functionality to the projection of full-fledged f-structures from chunk-based analyses of configurational languages.