Prototipo para el reconocimiento de entidades nombradas en el idioma Español

This paper propose a model based on Named Entities Recognition, using Conditional Random Fields (CRFs) for Spanish language. In this proposal different types of independent language features that provide local information. These features are useful in the identification and classification of named entities: people, organizations, places and others. Two annotated corpus, ANCORA and CoNLL2002. The were modified by adding some independent language features presented in this article. Our findings suggest that CRFs are a useful approach for Named Entity Recognition. The tests were conducted using perfrmance measures as: Precision, Recall, and F-Measure. Of the one hand CoNLL2002 corpus, Precision results were obtained between 79% and 86%, Recall between 64% and 79% and F-Measure between 73 and 80 and the other hand, ANCORA corpus was measured using Cross Validation Method. Precision results were obtained between 77% and 78%, Recall between 47% and 49%, and F-Measure 59%, average.