Simplification de phrases pour l'extraction de relations (Sentence Simplification for Relation Extraction) [in French]

Sentence simplification for relation extraction Machine learning based relation extraction requires large annotated corpora to take into account the variability in the expression of relations. To deal with this problem, we propose a method for simplifying sentences, i.e. for reducing the syntactic variability of the relations. Simplification requires the annotation of a small corpus, which will be automatically augmented. The process starts with the annotation of the simplification thanks to a CRF classifier, then the relation extraction, and lastly the automatic completion of the training corpus for the simplification through the results of the relation extraction. The first results we obtained for the task of relation extraction of the i2b2 2010 challenge are encouraging. MOTS-CLÉS : Extraction de relations, simplification de phrases, apprentissage automatique.