ECAI 2004 Conference Paper

[PDF] [full paper] [prev] [tofc] [next]

Voted Co-training for Bootstrapping Sense Classifiers

Rada Mihalcea

This paper introduces voted co-training, a bootstrapping method that combines co-training with majority voting, with the effect of smoothing the learning curves, and improving the average performance. Voted co-training was evaluated on a word sense classification problem, with significant improvements observed over basic co-training algorithms. Various optimal and empirical parameter selection methods for co-training are also investigated, with various degrees of error reduction.

Keywords: word sense disambiguation, semantics, natural language processing, bootstrapping, machine learning

Citation: Rada Mihalcea: Voted Co-training for Bootstrapping Sense Classifiers. In R.López de Mántaras and L.Saitta (eds.): ECAI2004, Proceedings of the 16th European Conference on Artificial Intelligence, IOS Press, Amsterdam, 2004, pp.515-519.

[prev] [tofc] [next]

ECAI-2004 is organised by the European Coordinating Committee for Artificial Intelligence (ECCAI) and hosted by the Universitat Politècnica de València on behalf of Asociación Española de Inteligencia Artificial (AEPIA) and Associació Catalana d'Intel-ligència Artificial (ACIA).