Summarizing Definition from Wikipedia Articles

摘要

Definitional questions are quite important,since users often want to get a brief overview of a specific topic.It is a more challenging task to answer definitional questions than factoid questions.Since Wikipedia provides a wealth of structural or semi-structural information which covers a large number of topics,such sources will benefit the generation of definitions.In this paper,we propose a method to summarize definition from multiple related Wikipedia articles.First,we introduce the Wikipedia concepts model to represent the semantic elements in Wikipedia articles.Second,we further utilize multiple related articles,rather than a single article,to generate definition.The experiment results on TREC-QA demonstrate the effectiveness of our proposed method.The Wikipedia concept model outperforms the word model.Introducing multiple related articles helps find more essential nuggets.