Une archive numérique francophone des articles de recherche en Traitement Automatique de la Langue.

Multi-Document Summarization: Methodologies and Evaluations

Gees C. Stein, Amit Bagga, G. Bowden Wise

Abstract : This paper describes a system for the summarization of multiple documents. The system produces multi-document summaries using clustering techniques to identify common themes across the set of documents. For each theme, the system identifies representative passages that are included in the final summary. We also describe a methodology for evaluation of our system which is based upon a question answering task. Results of our evaluation are also presented.