Research on Brazilian Portuguese teaching as L1 has long advocated in favor
of grounding the study of grammar on the texts to which a given linguistic
structure is highly relevant (Geraldi 1991; Mendonça 2006). Exemplar
applications of this perspective include teaching imperatives using cooking
recipes or user’s manuals. However, work on analyzing teaching manuals for
Brazilian Portuguese, both by academia and governmental agencies, has
shown that kind of exemplar applications are still rare, that is: although the
work with some linguistic structures tend to be adequately grounded on the
genres to which they are relevant, manuals approach the majority of grammar
using sentences removed from their context of actual usage (Neves 2002).
One of the reasons leading to this scenario is related to the lack of easily
accessible systematic descriptions of the grammatical properties of full texts
in Brazilian Portuguese. The project being presented here, hence, aims to
build a freely available computational resource in which texts annotated for
the grammatical structures instantiated in them can be searched and
extracted for language teaching purposes. Annotation follows the
lexicographic and constructional annotations guidelines of FrameNet Brasil
(Torrent et al. 2014), a computational resource for Brazilian Portuguese
covering both lexical and grammatical properties of this language. In this
paper, we present a case study based on the annotation of travel guides,
showing which kinds of grammatical structures are more relevant to this
genre. We also discuss the potential of FrameNet as a means to identify
genre-relevant grammatical structures and also to show how they pair to the
semantic structure of the text.
