Semantic Representation of Ellipsis in the Prague Dependency Treebanks

The 2014 Conference on Computational Linguistics and Speech Processing ROCLING 2014, pp. 125-138  The Association for Computational Linguistics and Chinese Language Processing Semantic Representation of Ellipsis in the Prague Dependency Treebanks Marie Mikulová Abstract This article answers the question what is and what is not ellipsis and specifies criteria for identification of elliptical sentences. It reports on an analysis of types of ellipsis from the point of view of semantic representation of sentences. It does not deal with conditions and causes of the constitution of elliptical positions in sentences (when and why is it possible to omit something in a sentence) but it focuses exclusively on the identification of elliptical positions (if there is something omitted and what) and on their semantic representation in a treebank, specifically on their representation on the deep syntactic level of the Prague Dependency Treebanks. The theoretical frame of the approach to ellipsis presented in this article is dependency grammar. Keywords: Ellipsis, Semantic Annotation, the Prague Dependency Treebanks. 1. Introduction The analysis of well-formed sentences, some of whose constituents are missing, has been of central concern to computational linguists at least since the beginnings of the work in formal grammar. There has been a considerable amount of research on ellipsis from a variety of perspectives. Different approaches to an explanation of the procedures involved in assigning representations to sentences containing deletions have been developed (see, for example, Berman – Hestsvik, 1992 and Lappin – Benmamoun, 1999) but they have been mostly designed within the framework of constituency grammar. The theoretical frame of our approach to ellipsis is dependency grammar. The dependency-based approach offers a totally different perspective on ellipsis. The constituency-based approach assigns more empty positions. The so-called gaps are assigned particularly when two constituents cannot be bracketed because they do not occur one next to the other in the surface form of a sentence. This discontinuity of two constituents does not impede the construction of a dependency tree. Only those gaps are perceived as ellipses when  Charles University in Prague, Faculty of Mathematics and Physics, Institute of Formal and Applied Linguistics, Czech Republic; E-mail: mikulova@ufal.mff.cuni.cz. 125 one of the constituents is not expressed at all in the surface form of the sentence. Thus, only a small part of the gaps (identified by constituency-based approaches) overlaps with the types of ellipses that have been defined by means of dependency syntax. This article answers the question what is and what is not ellipsis in dependency grammar. It reports on an analysis of types of ellipsis from the point of view of semantic representation of sentences in a dependency treebank. We identify a boundary between the grammatical ellipses on the one hand and the accidental omissions on the other. We will distinguish morphological, surface and deep syntactic features of ellipses. 2. What is Ellipsis? In attempting to characterize elided sentences in natural language we are faced with the problem of explaining how speakers or recipients are able to represent and interpret linguistic objects which, at …

Content truncated to 500 words