Experiments on authorship attribution by intertextual distance in english*

Abstract

How can it be said that texts are “near to” or “distant from” one another? Are different texts by a single author more similar than texts by different authors? To answer these questions, a method is proposed by calculating intertextual distance. A blind test and some additional experiments show that this calculation offers an interesting tool for non-traditional authorship attribution.

Acknowlegments

The author is grateful to Cyril Labbé (Grenoble I University) who wrote the software with him, to Gerard Ledger for providing the two Corpora Oxquarry and organising the blind-test, and to Thomas Merriam: he had the idea of applying our methods on English texts, he gave me a continuous support and most helpful comments and advice, he carefully edited the first version of this article.

Notes

¹1A: Hardy (Jude); 1I: Orczy (Scarlet_P); 1T: Conrad (Almayer); 1U: Orczy (Elusive_P); 1V Conrad (Lord_Jim).

²These three excerpts come from Catrionae by Stevenson (Appendix 1).

³In fact, these four solutions are correct: 1V, 1U, sets 9 and 11 are by Orczy; Hardy is the author of Set 4 and 5; Conrad is the one of 1T and set 10 (Appendix 1).

⁴Here ends the text sent to G. Ledger in February 2004 in order to present him the results of the “blind-test”.