Using semantic relatedness and word sense disambiguation for (CL)IR

Abstract

In this paper we report the experiments for the CLEF 2009 Robust-WSD task, both for the monolingual (English) and the bilingual (Spanish to English) subtasks. Our main experimentation strategy consisted of expanding and translating the documents, based on the related concepts of the documents. For that purpose we applied a state-of-the art semantic relatedness method based on WordNet. The relatedness measure was used with and without WSD information. Even though we obtained positive results in our training and development datasets, we did not manage to improve over the baseline in the monolingual case. The improvement over the baseline in the bilingual case is marginal. We plan further work on this technique, which has attained positive results in the passage retrieval for question answering task at CLEF (ResPubliQA). © 2010 Springer-Verlag Berlin Heidelberg.