La reconnaissance d’entités nommées et la granularité des ressources dictionnairiques

Abstract

In this article we briefl y report on different architectures of the computer systems dedicated to the identifi cation of named entities in texts, such as person name, location, organization and date. These are able to identify named entities with a precision and a recall rate exceeding 0.9. We then discuss two types of named entities designating people and show the importance of the granularity of semantic descriptions associated with each entry in very large coverage electronic dictionaries contained in the named entity recognition system.