Abstract EN:

This thesis explores ways to define automated coreference resolution systems by using structured machine leaming techniques. We design supervised models that leam to build coreference clusters from raw text: our main objective is to get model able to process documents globally, in a structurel fashion, to ensure coherent outputs. Our models are trained and evaluated on the English part of die CoNLL-2012 Shared Task annotated corpus with standard metrics. We carry out detailed comparisons of different settings so as to refine our models and design a complete end-to-end coreference resolver. Specifically, we first carry out a preliminary work on improving the way features are employed by linear models for classification: we extend existing work on separating different types of mention pairs to define more accurate classifiers of coreference links. We then define varions structured models based on latent trees to learn to build clusters globally, and not only from die predictions of a mention pair classifier. We study different latent representations (varions shapes and sparsity) and show empirically that die best suited structure is some restricted class of trees related to the best-first rule for selecting coreference links. We further improve this latent representation by integrating anaphoricity modelling jointly with coreference, designing a global (structured at the document level) and joint model outperforming existing models on gold mentions evaluation. We finally design a complete end-to-end resolver and evaluate the improvement obtained by our new modela on detected mentions, a more realistic setting for coreference resolution.