The Elements of Automatic Summarization

The Elements of Automatic SummarizationbyDaniel Jacob GillickDoctor of Philosophy in Computer ScienceUniversity of California, BerkeleyProfessor Nelson Morgan, Chair This thesis is about automatic summarization, with experimental results on multi-document news topics: how to choose a series of sentences that best represents a col-lection of articles about one topic. I describe prior work and my own improvementson each component of a summarization system, including preprocessing, sentencevaluation, sentence selection and compression, sentence ordering, and evaluation ofsummaries. The centerpiece of this work is an objective function for summariza-tion that I call "maximum coverage". The intuition is that a good summary coversas many possible important facts or concepts in the original documents. It turnsout that this objective, while computationally intractable in general, can be solvedefficiently for medium-sized problems and has reasonably good fast approximate so-lutions. Most importantly, the use of an objective function marks a departure fromprevious algorithmic approaches to summarization.