Word Sense Induction in the Arabic Language: A Self-Term Expansion Based Approach
David Pinto, Paolo Rosso, Yassine Benajiba, Anas Ahachad, Héctor Jiménez-Salazar · 2007
Abstract. The aim of the word sense induction/discrimination task of natural language processing is to discover the sense associated to each instance of a given ambiguous word. In this paper we present an approach based on clustering of a self-expanded version of the original dataset in order to tackle this particular problem. The self-expansion technique substitutes every term of the original corpus with a set of co-related terms which is calculated by means of pointwise mutual information. Our proposal which was tested for the English language shows a good performance for the Arabic language too, highlighting its languageindependent characteristic.