Please use this identifier to cite or link to this item:
|Scopus||Web of Science®||Altmetric|
|Title:||Advanced text authorship detection methods and their application to biblical texts|
|Citation:||Proceedings of SPIE, 2006 / Bender, A. (ed./s), vol.6039, pp.60390J-1-60390J-13|
|Series/Report no.:||Proceedings of SPIE--The International Society for Optical Engineering ; 6039|
|Conference Name:||SPIE Microelectronics, MEMS, and Nanotechnology (11 Dec 2005 - 14 Dec 2005 : Brisbane, Australia)|
|Talis Putnins, Domenic J. Signoriello, Samant Jain, Matthew J. Berryman, and Derek Abbott|
|Abstract:||Authorship attribution has a range of applications in a growing number of fields such as forensic evidence, plagiarism detection, email filtering, and web information management. In this study, three attribution techniques are extended, tested on a corpus of English texts, and applied to a book in the New Testament of disputed authorship. The word recurrence interval based method compares standard deviations of the number of words between successive occurrences of a keyword both graphically and with chi-squared tests. The trigram Markov method compares the probabilities of the occurrence of words conditional on the preceding two words to determine the similarity between texts. The third method extracts stylometric measures such as the frequency of occurrence of function words and from these constructs text classification models using multiple discriminant analysis. The effectiveness of these techniques is compared. The accuracy of the results obtained by some of these extended methods is higher than many of the current state of the art approaches. Statistical evidence is presented about the authorship of the selected book from the New Testament.|
|Description:||© 2006 COPYRIGHT SPIE--The International Society for Optical Engineering|
|Appears in Collections:||Aurora harvest 2|
Electrical and Electronic Engineering publications
Files in This Item:
There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.