Filling the gap: inserting an artificial constituent where a subject is omitted in Portuguese

Abstract

This paper reports the first efforts to insert null elements to represent omitted subjects in Portuguese. Our aim is to fill some gaps in the syntactic structure in order to facilitate the assignment of semantic role labels and thus provide a better training corpus for SRL classifiers. The main advantage of inserting such null elements is to reduce data sparsity, as all the verbal clauses become similar in what concerns the presence of explicit subjects. The results show a better precision in the insertion of null elements related to subjects of verbs inflected in the first person, both singular and plural.