The DUC 2002 evaluation revealed numerous language-processing challenges that impact text summarization. The authors examine the techniques used in a multidocument summarization system they developed and its performance at DUC 2002. They also discuss the need for regularization of human summaries.
