Evaluating Summaries and Answers: Two Sides of the Same Coin?

摘要

This paper discusses the convergencebetween question answering and multidocumentsummarization, pointing outimplications and opportunities for knowledgetransfer in both directions. As acase study in one direction, we discussthe recent development of an automaticmethod for evaluating definition questionsbased on n-gram overlap, a commonlyusedtechnique in summarization evaluation.In the other direction, the move towardstopic-oriented summaries requiresan understanding of relevance and topicality,issues which have received attentionin the question answering literature.It is our opinion that question answeringand multi-document summarization representtwo complementary approaches to thesame problem of satisfying complex userinformation needs. Although this pointsto many exciting opportunities for systembuilding,here we primarily focus on implicationsfor system evaluation.