Computational analysis of adjuncts in ASD-STE100 for the NLP parser ARTEMIS

Resumo

Texts written in Simplified Technical English present semantic and syntactic restrictions with respect to standard English that should be considered in the implementation of natural language processing systems dealing with controlled natural languages. This paper explores the syntax and semantics of adjuncts in a controlled natural language, namely, the Simplified Technical English (ASD-STE100) used in the Airbus corpus with a view to observing how the peculiarities identified in the behaviour of Airbus adverbials should be reflected in the computational treatment of these constituents. Thus, our main aim is to contribute to the computational implementation of the prototype ARTEMIS by designing the parsing rules and the catalogue of feature-bearing matrixes that encode the grammatical constraints of Airbus adjuncts. The parsing rules reflect the positional preferences of the various peripheral types and capture the semantic variability of adjuncts in the corpus. In addition, they provide a weight factor that predicts the scale of markedness of these constituents with respect to the different positions that they occupy in the clause. On the whole, these properties offer a precise description of the syntactic features of adverbials that will facilitate their automatic processing.