On the Use of Supervised Learning Method for Authorship Attribution

Abstract

In this paper we investigate the use of a supervised learning method for theauthorship attribution that is for the identification of the author of a text. Wesuggest a new, simple and efficient method, which is merely based on counting thenumber of repetitions of each alphabetic letter in the text, instead of using thetraditional classification properties; such as the contents of the text and style of theauthor; which falls into four feature categories: lexical, syntactic, structural, andcontent-specific. Furthermore, we apply a spherical classification method.We apply the proposed technique to the work of two Italian writers, DanteAlighieri and Brunetto Latini. With almost high reliability, the spherical classifierproved its ability to discriminate between the selected authors.Finally the results are compared with those obtained by means of a standardSupport Vector Machine classifier.