Cut-and-paste text summarization

TL;DR: This thesis presents a cut-and-paste approach to addressing the text generation problem in domain-independent, single-document summarization, and built a large-scale, reusable lexicon by combining multiple, heterogeneous resources.

Abstract: Automatic text summarization provides a concise summary for a document. In this thesis, we present a cut-and-paste approach to addressing the text generation problem in domain-independent, single-document summarization. We found that professional abstractors often reuse the text in an original document for producing the text in a summary. But rather than simply extracting the original text, as in most existing automatic summarizers, humans often edit the extracted sentences. We call such editing operations “revision operations”. Our summarizer simulates two revision operations that are frequently used by humans: sentence reduction and sentence combination. Sentence reduction removes inessential phrases from sentences and sentence combination merges sentences and phrases together. The sentence reduction algorithm we propose relies on multiple sources of knowledge to decide when it is appropriate to delete a phrase from a sentence, including linguistic knowledge, probabilities trained from corpus examples, and context information. The sentence combination module relies on a set of rules to decide how to combine sentences and phrases and when to combine them. Sentence reduction aims to improve the conciseness of generated summaries and sentence combination aims to improve the coherence of generated summaries. We call this approach “cut-and-paste” since it produces summaries by excerpting and combining sentences and phrases from original documents, unlike the extraction technique which produces summaries by simply extracting sentences or passages. Our work also includes a Hidden Markov Model based sentence decomposition program which analyzes human-written summaries. The decomposition program identifies where the phrases of a summary originate in the original document, producing an aligned corpus of summaries and articles that we use to train and evaluate the summarizer. We also built a large-scale, reusable lexicon by combining multiple, heterogeneous resources. The lexicon contains lexical, syntactic, and semantic knowledge. It can be used in many applications.