Contribution de la métrique à la stylométrie

In this paper, we present the results of our first attempts to use metrical informations for authorship attribution. 58 plays in verse by Corneille, Racine and Molière were systematically analyzed regarding syllabic, morpho- syntactic and stress structure, using the Metrometer, a Natural Language Processing tool aimed at the analysis of French classical verse. Our main finding are that (i) Corneille and Racine are very consistent in their metrical use of words, Molière being quite less so; (ii) using statistical language modeling tools, it is in fact possible to build syllable-based model which can effectively discriminate genre and authors.