Improving speculative language recognition incorporating expert knowledge

Guillermo Moncecchi (1, 2) , Jean-Luc Minel (1) , Dina Wonsever (2)

1 MoDyCo - Modèles, Dynamiques, Corpus
2 INCO - Instituto de Computacion [Montevideo]

In this paper we present a methodology to improve a supervised classifier analyzing classification errors and incorporating expert knowledge to the learning process, without hurting classifier performance. We implement and evaluate the methodology on the tasks of hedge cue identification and scope resolution.

Mots clés: supervised classification; hedge cue identification

Domaines: Linguistique

Guillermo Moncecchi, Jean-Luc Minel, Dina Wonsever. Improving speculative language recognition incorporating expert knowledge. Roser Morante, Caroline Sporleder. ExProM 2012, Jul 2012, Jeju, South Korea. Association for Computational Linguistics (ACL), pp.37-46, 2012. ⟨halshs-00721590⟩