Producing Intelligent Telegraphic Text Reduction to provide an Audio Scanning Service for the Blind

摘要

Text summarization is usually taken to mean producing a shorter version of an original document by retaining the most salient parts of the original text. Two approaches have been favored: selecting high content-bearing sentences influenced by positional constraints, and performing domain-dependent information extraction which fills a template from which a summary can be glossed. This paper presents a third approach to text reduction, producing shortened telegraphic versions of all input sentences. This approach shares the domain-independence advantage of the sentence selection and the one-pass advantage of information extraction template filling. We argue that this type of text reduction has some practical applications, and present the case of creating an audio scanner for the blind.