Sintaxe formal de línguas não configuracionais num ambiente computacional: o caso do Latim

Abstract

In this paper, we present a formal analysis of some aspects of Latin syntax, pointing out difficulties prototypical non-configurational languages like this represent to a computational treatment within the DCG formalism in its classical version. We propose an alternative notation that exploits SWI-Prolog's advanced features to extend the expressive power of DCG. In this more efficient format, a generative Latin grammar fragment is built that copes with free order of constituents in sentences with a variety of divalent and trivalent verbs exhibiting both canonical and non-canonical case and argument linking patterns. In our proposal, we resort to two general rules formalizing case checking and agreement, so that only two phrase structure rules are needed to generate sentences with two-place and three-place verbs. By contrast, hundreds of rules are necessary in the traditional DCG format to model the same corpus. On the other hand, thanks to Prolog built-in parsing algorithm, our grammar fragment not only recognizes and analyses sentences with both expletive and thematic null subjects, but also does the mapping from morphosyntactic to semantic structure. By this means, it is capable of building different representations of ambiguous sentences resulting from more than one possible syntactic-semantic mapping, expliciting the different ways how the verb's argument structure is linked to syntactic arguments in the sentence.