Ensemble Learning of Economic Taxonomy Relations from Modern Greek Corpora

This paper proposes the use of ensemble learning for the identification of taxonomic relations between Modern Greek economic terms. Unlike previous approaches, apart from is-a and part-of relations, the present work deals also with relation types that are characteristic of the economic domain. Semantic and syntactic information governing the term pairs is encoded in a novel feature-vector representation. Ensemble learning helps overcome the problem of performance instability and leads to more accurate