Boosting Grammatical Inference with Confidence Oracles
Jean-Christophe Janodet - EURISE-Faculty of Sciences-University of Saint-Etienne Richard Nock - GRIMAAG-French West Indies and Guyana University Marc Sebban - EURISE-Faculty of Sciences-University of Saint-Etienne Henri-Maxime Suchier - EURISE-Faculty of Sciences-University of Saint-Etienne
In this paper we focus on the adaptation of boosting to grammatical inference.We aim at improving the performances of state merging algorithms in thepresence of noisy data by using, in the update rule, additional informationprovided by an oracle. This strategy requires the construction of a newweighting scheme that takes into account the confidence in the labels of theexamples. We prove that our new framework preserves the theoretical propertiesof boosting. Using the state merging algorithm RPNI*, we describe anexperimental study on various datasets, showing a dramatic improvement ofperformances.