An Overview of the Traditional Authorship Attribution Subtask.
Patrick Juola · 2012
Abstract This paper describes the Traditional Authorship Attribution subtask of the PAN/CLEF 2012 workshop. As a followup to our subtask at PAN/CLEF 2011 (Amsterdam), we established a new corpus for analysis for 2012 (Rome). The new corpus differed in several ways from the previous subtask: – Both the number and size of documents were decreased – The documents were taken from a different genre (fiction, represented by the Feedbooks.com site) – The documents were no longer marked up extensively – A new sub-sub-task was added: Authorship clustering. In this new problem (related to Òintrinsic plagiarismÓ) participants were given a text of mixed authorship and asked to determine which paragraphs came from which au-thors. The resulting corpus consisted of eight problems, including three closed-class authorship attribution problems, three open-class (the set of correct answers in-cluded Ònone of the aboveÓ), and two clustering problems. Twenty-five teams participated in this subtask from many different parts of the world. Detailed re-sults are available on the Web at pan.webis.de and will be discussed in detail at the PAN/CLEF 2012 meeting in September. 1