Authorship attribution for electronic documents

Abstract

Forensic analysis of questioned electronic documents is difficult because the nature of the documents eliminates many kinds of informative differences. Recent work in authorship attribution demonstrates the practicality of analyzing documents based on authorial style, but the state of the art is confusing. Analyses are difficult to apply, little is known about error types and rates, and no best practices are available. This paper discusses efforts to address these issues, partly through the development of a systematic testbed for multilingual, multigenre authorship attribution accuracy, and partly through the development and concurrent analysis of a uniform and portable software tool that applies multiple methods to analyze electronic documents for authorship based on authorial style.