A New Stylometric Technique for the Invesigation of the True Authorship of the Early Works of Shakespeare

Abstract

This thesis follows the development of a new stylometric technique which can be used to investigate works of literature of questionable authorship. A particular dispute is examined concerning the true authorship of three plays commonly credited to William Shakespeare. The works are the two historical plays Henry VI parts 2 and 3 and the Roman tragedy Titus Andronicus. The three works have been argued to have been greatly influenced by the playwright Christopher Marlowe. In a problem which is prone to high noise and limited data availability, intelligent feature selection is an essential part of the classification process. This should lead to the identification of the most promising inputs to use in discriminating between the two authors. The use of both linear classifier and non-linear Neural Network based classifier models are investigated and their results compared.