Why glue a Donkey to an f-structure when you can constrain and bind it instead?

Miltiadis Kokkonidis

Abstract

Proceedings of LFG05; CSLI Publications On-line

The semantic treatment of anaphora using lamda-DRT with Glue presented combines the strengths of both by assigning to each the task where it arguably fares best: Glue composes meanings and DRT deals with anaphoric resolution. Key to this approach is a simple first-order system for lamda-DRT that allows LFG syntactic constraints to be transferred into the dynamic representation language. This parallels the transfer of such constraints into Glue types. Whereas approaches treating anaphora using Glue context management take advantage of this transfer, an earlier approach also leaving the treatment of anaphora to a compositional variant of DRT failed to account for syntactically-motivated anaphoric resolution constraints. On the other hand, the existing Glue context management approaches come not only at the cost of coupling context management with meaning composition, but also at the additional cost of the various remedies to the problems this uneasy cohabitation results in. The best of these approaches and the one presented here are currently very similar with respect to the range of phenomena they can correctly account for, but there are reasons to believe the latter is more scaleable.