A Profile-Based Method for Authorship Verification

Abstract

Abstract. Authorship verification is one of the most challenging tasks in stylebased text categorization. Given a set of documents, all by the same author, and another document of unknown authorship the question is whether or not the latter is also by that author. Recently, in the framework of the PAN-2013 evaluation lab, a competition in authorship verification was organized and the vast majority of submitted approaches, including the best performing models, followed the instance-based paradigm where each text sample by one author is treated separately. In this paper, we show that the profile-based paradigm (where all samples by one author are treated cumulatively) can be very effective surpassing the performance of PAN-2013 winners without using any information from external sources. The proposed approach is fully-trainable and we demonstrate an appropriate tuning of parameter settings for PAN-2013 corpora achieving accurate answers especially when the cost of false negatives is high.

Authorship Attribution Using Text Distortion

“…In this study, we use the authorship verification approach proposed by Potha and Stamatatos (2014). In more detail, this is a profile-based approach meaning that first it concatenates all available known documents and then it extracts a single representation from the resulting document (Stamatatos, 2009).…”

Literary writing style recognition via a minimal spanning tree-based approach

“…The profile-based paradigm is investigated in Potha and Stamatatos (2014) . According to this paradigm, all text samples of a given author are concatenated into a single document.…”

Authenticating the writings of Julius Caesar

“…The first verification system (termed O1 here) used here was seminally introduced by Kjell et al (Kešelj et al, 2003) and was subsequently refined (Potha & Stamatatos, 2014;Kestemont et al, 2011;Stamatatos, 2009a). O1 resorts to the direct (or 'first order') calculation of a distance metric between a target author's stylistic profile in a given problem, and the unknown text.…”

“…O1 resorts to the direct (or 'first order') calculation of a distance metric between a target author's stylistic profile in a given problem, and the unknown text. Following (Potha & Stamatatos, 2014;Koppel & Seidman, 2013), we define an author's profile here as the mean centroid of the known 145 document vectors for that author (i.e. we average an author's score for a particular term across all training texts).…”

“…To verify whether the unknown document was written by the target score to probability scores in the 0-1 range, they are scaled using the set of all non-zero pairwise scores which can obtained between the known documents in a problem set, before their positive complement is taken (Potha & Stamatatos, 2014). While O1 has so far primarily been used with the cng metric, it can 160 also be used with the other distance metrics introduced below.…”