Etude de l'impact du regroupement automatique de phrases sur un système de résumé multi-documents

This paper introduces the experiences we led in order to evaluate the impact of a sentence unsupervised clustering algorithm on a multi-document summarization system. We compared this system to two others which implement classic summarization methods. We also evaluated the impact of different sentence similarity measures on the quality of the summaries. The experiences show that a clustering prior to sentence selection does improve the quality of the summaries.