|
Title
|
|
|
|
Overview of the cross-domain authorship attribution task at {PAN} 2019
| |
|
Author
|
|
|
|
| |
|
Abstract
|
|
|
|
| Authorship identification remains a highly topical research problem in computational text analysis, with many relevant applications in contemporary society and industry. In this edition of PAN, we focus on authorship attribution, where the task is to attribute an unknown text to a previously seen candidate author. Like in the previous edition we continue to work with fanfiction texts (in four Indo-European languages), written by non-professional authors in a crossdomain setting: the unknown texts belong to a different domain than the training material that is available for the candidate authors. An important novelty of this year’s setup is the focus on open-set attribution, meaning that the test texts contain writing samples by previously unseen authors. For these, systems must consequently refrain from an attribution. We received altogether 12 submissions for this task, which we critically assess in this paper. We provide a detailed comparison of these approaches, including three generic baselines. |
| |
|
Language
|
|
|
|
English
| |
|
Source (book)
|
|
|
|
Working Notes of CLEF 2019 - Conference and Labs of the Evaluation Forum, Lugano, Switzerland, September 9-12, 2019
| |
|
Publication
|
|
|
|
2019
| |
|
Volume/pages
|
|
|
|
p. 1-15
| |
|
Full text (open access)
|
|
|
|
| |
|