Section mixture models for scientific document summarization

A system for summarization of scientific and structured documents that has three components: section mixture models are used for estimation of the weights of terms; a hypothesis test to select a subset of these terms; and a sentence extractor based on techniques for combinatorial optimization is presented.