The RST Spanish Treebank On-line Interface

Proceedings of Recent Advances in Natural Language Processing, pages 698–703, Hissar, Bulgaria, 12-14 September 2011. The RST Spanish Treebank On-line Interface Iria da Cunha Instituto Universitario de Lingüística Aplicada (UPF) iria.dacunha@upf.edu Juan-Manuel Torres-Moreno Laboratoire Informatique d’Avignon, Universidad Nacional Autónoma de México, École Polytechnique de Montréal juan-manuel.torres @univ-avignon.fr Gerardo Sierra Universidad Nacional Autónoma de México gsierram@iingen. unam.mx Luis-Adrián Cabrera-Diego Universidad Nacional Autónoma de México LCabreraD@ iingen.unam.mx Brenda-Gabriela Castro-Rolón Universidad Nacional Autónoma de México BCastroR@iingen.unam.mx Juan-Miguel Rolland- Bartilotti Universidad Nacional Autónoma de México jrollandb@ iingen.unam.mx Abstract In this article, we present the on-line interface that we have developed for the RST Spanish Treebank, the first corpus including Spanish texts annotated with rhetorical relations. This interface allows users to consult or download the texts and their corresponding annotations. In addition, it allows carrying out several tasks over a selected subcorpus: searching statistics in terms of words, rhetorical relations and Elementary Discourse Units (EDUs), and extracting information, in terms of texts passages marked with rhetorical relations (ex. Result, Cause or Background), which users may select. 1 Introduction According to Hovy (2010), there are 7 core questions in corpus’ design: selecting a corpus, instantiating the theory, designing the interface, selecting and training the annotators, designing and managing the annotation procedure, validating results, and delivering and maintaining the product. All these points are really relevant when compiling a corpus. However, we consider that usually one of them is underestimated: the interface design. When Hovy (2010) mentions this aspect, he mainly refers to the annotation interface. We think that the annotation interface is important but, as well, that, if there is an available annotation interface suitable for the purposes of a corpus project, it can be used. Nevertheless, we consider that an interface allowing users to consult or download the corpus’ texts, and even carrying out searches (both statistics and linguistics) over a selected subcorpus, is really useful and necessary. Compiling and annotating an adequate corpus is not a trivial task; it implies lots of people, resources, time and effort. Thus, we consider that it is important to develop a friendly and useful interface to be able to exploit the created corpus, and transform it into a most accessible resource. Therefore, in this article, we present the on-line interface that we have developed in order to include the RST Spanish Treebank (da Cunha et al., 2011). The RST Spanish Treebank is the first corpus including Spanish texts annotated with rhetorical relations of the Rhetorical Structure Theory (RST) by Mann and Thompson (1988). It contains texts from nine specialized domains (Astrophysics, Earthquake Engineering, Economy, Law, Linguistics, Mathematics, Medicine, Psychology and Sexuality). It includes 52,746 words, 267 texts, 2,256 sentences and 3,349 discourse segments. The segmentation criteria are similar to those employed by da Cunha et al. (2011). Each text was tagged by 1 person, from a team of 10 RST expert annotators. There is a 31% of the corpus double-annotated. The corpus is not annotated with syntactic structure, although we are conscious this would be interesting. The corpus …

Content truncated to 500 words