Joint Anaphoricity Detection and Coreference Resolution with Constrained Latent StructuresReport as inadecuate

Joint Anaphoricity Detection and Coreference Resolution with Constrained Latent Structures - Download this document for free, or read online. Document in PDF available to download.

1 ALPAGE - Analyse Linguistique Profonde à Grande Echelle ; Large-scale deep linguistic processing Inria Paris-Rocquencourt, UPD7 - Université Paris Diderot - Paris 7 2 MAGNET - Machine Learning in Information Networks Inria Lille - Nord Europe, CRIStAL - Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189

Abstract : This paper introduces a new structured model for learninganaphoricity detection and coreference resolution in a jointfashion. Specifically, we use a latent tree to represent the fullcoreference and anaphoric structure of a document at a globallevel, and we jointly learn the parameters of the two modelsusing a version of the structured perceptron algorithm.Our joint structured model is further refined by the use ofpairwise constraints which help the model to capture accuratelycertain patterns of coreference. Our experiments on theCoNLL-2012 English datasets show large improvements inboth coreference resolution and anaphoricity detection, comparedto various competing architectures. Our best coreferencesystem obtains a CoNLL score of 81:97 on gold mentions,which is to date the best score reported on this setting.

Keywords : coreference resolution anaphoricity structure prediction structured perceptron joint learning

Author: Emmanuel Lassalle - Pascal Denis -



Related documents