Building a Resource for Evaluating the Importance of Sentences
Barry Schiffman (Columbia University, Department of Computer Science New York, N.Y. 10027, U.S.A.)
WP5: Components & Systems
This paper will introduce a new lexical resource for measuring the importance of short segments of text, such as sentences. The resource, a list of words compiled automatically from a large background corpus of news articles, can provide evidence that a text segment is globally important, that is intrinsically interesting, not only interesting in relation to a specied topic or set of documents.