The application of linguistic processing to automatic abstract generation

Johnson, Frances C., Paice, Christopher D., Black, William J. and Neal, A.P. (1997) The application of linguistic processing to automatic abstract generation. In: Readings in information retrieval. Morgan Kauffman, pp. 538-553. ISBN 9781558604544

Preview

Accepted Version Download (185kB) | Preview

Abstract

One approach to the problem of generating abstracts by computer is to extract from a source text those sentences which give a strong indication of the central subject matter and findings of the paper. Not surprisingly, concatenations of extracted sentences show a lack of cohesion, due partly to the frequent occurrence of anaphoric references. This paper describes the text processing which was necessary to identify these anaphors so that they may be utilised in the enhancement of the sentence selection criteria. It is assumed that sentences which contain non-anaphoric nounphrases and introduce key concepts into the text are worthy of inclusion in an abstract. The results suggest that the key concepts are indeed identified but the abstracts are too long. Further recommendations are made to continue this work in abstracting which makes use of text structure.