Extraction automatique de termes-clés : Comparaison des méthodes non supervisées de la littérature

This article presents a state of the art and a co mparison of unsupervised methods for automatic keywords extraction from documents. We evaluate several methods from the literature on two sets of documents by comparing the keywords extracted and those initially associated with documents. We found that the best me thod (the one that retrieves keywords the closest to the authors’ keywords) is based on TF -IDF. MOTS-CLÉS : extraction automatique de termes-clés, méthodes n on supervisées, représentation