Evaluating lexical resources for WSD

By November 17, 2016,
AuthorM. Taulé, M. García, N. Artigas, M. A. Martí
TitleEvaluating lexical resources for WSD
AbstractIn this paper we present a methodology developed for evaluating the quality of the dictionaries for Word Sense Disambiguation (WSD) tasks. This methodology is based on the parallel tagging of a single corpus with three lexical sources (with different characteristics) and three annotators. As far as the corpus tagging is concerned, strict criteria have been established regarding the lexicographic training of the annotators as well as the process itself, which has been developed under the guidance of a handbook specially elaborated for this task. The evaluation of the annotators' agreement degree has been automatically drawn by means of previously defined agreement measurements (total agreement, partial agreement, minimum agreement, and disagreement). As a result, two new tested resources have been obtained or the treatment of Spanish language: a lexical source, MiniDir.2.1, suitable for WSD tasks, and a semantically tagged Spanish corpus, MiniCors.
SessionComputational Lexicography and Lexicology
author = {M. Taulé, M. García, N. Artigas, M. A. Martí},
title = {Evaluating lexical resources for WSD },
pages = {197-206},
booktitle = {Proceedings of the 11th EURALEX International Congress},
year = {2004},
month = {july},
date = {6-10},
address = {Lorient, France},
editor = {Geoffrey Williams and Sandra Vessier},
publisher = {Université de Bretagne-Sud, Faculté des lettres et des sciences humaines},
isbn = {29-52245-70-3},