Towards an ISO Standard for the Annotation of Quantification

Towards an ISO Standard for the Annotation of Quantification Harry Bunt1, Kiyong Lee3, and James Pustejovsky2 1Tilburg Center for Communication and Cognition, Tilburg University, The Netherlands 2Brandeis University, Waltham, MA, USA 3Korea University, Seoul, Korea bunt@uvt.nl, pustejovsky@cs.brandeis.edu, ikiyong@gmail.com Abstract This paper presents an approach to the annotation of quantification that is being developed in preparation of the specification of a quantification annotation scheme, as part of an effort of the International Organisation for Standardisation ISO to define interoperable semantic annotation schemes. The paper focuses on the theoretical basis for an ISO standard annotation scheme in this area. It is argued that the combination of Generalized Quantifier Theory, neo-Davidsonian event-bases semantics, Discourse Representation Theory, and the ISO Principles of semantic annotation forms a powerful and solid foundation for defining annotations of quantification phenomena with an abstract and a concrete syntax and a compositional semantics. The coverage of the proposed annotation scheme includes both count and mass NP quantifiers, as well as NPs with syntactically and semantically complex heads with internal quantification and scoping structures, such as inverse linking by prepositional phrases and relative clauses. Keywords: semantic annotation, annotation standaard, quantification 1. Introduction Quantification occurs in every sentence of written text or spoken discourse. This is because the application of a pred- icate to one or more sets of objects gives rise to questions of relative scope, of cardinality, and of the distribution (or ‘distributivity’) of the predicate over the sets of arguments. Dealing with these questions is of crucial importance for understanding, translating, or responding to sentences, and for correct information extraction and question answering from natural language text. Example sentence (1a) illus- trates the influence of the relative scopes of quantifications on the information that an utterance contains (how many papers weere read altogether?); (1b) illustrates the impor- tance of the collective or individual distribution of a predi- cate over a set of arguments (assuming that a certain piano was carried collectively by certain men, who individually had a beer); and (1c) shows that a quantified relation may involve parts of individuals, such as pizza halves. (1) a. All the students read two papers on quantification. b. The men had a beer before carrying the piano up- stairs. c. The three boys ate four and a half pizzas. The International Organisation for Standardisation ISO is pursuing the establishment of standards for linguistic an- notation in general, and semantic annotation in particular, in view of the importance of annotated language resources, both for empirically-based linguistic research and for de- veloping NLP applications, In order to be widely appli- cable across theories and platforms, annotation standards should on the one hand be theory-neutral, but they should on the other hand also take established theoretical insights into account. This paper outlines an ISO standard anno- tation scheme under development for quantification that builds on logical and linguistic theories, notably on the the- ory of generalized quantifiers, on neo-Davidsonian event- based semantics, and on Discourse Representation Theory. in combination with established principles for semantic …

Content truncated to 500 words