Stylometric authorship attribution techniques and analysis for collaborative platforms
Edwin Dauber, Rachel Greenstadt, Spiros Mancoridis · 2020
Stylometry is defined as the study of writing style. While it has traditionally been studied in natural language, recent advances have also seen it applied to source code. There are many topics of interest in stylometry research, but the most common is authorship attribution - attempting to determine the author of a currently anonymous or disputed document. Most of this research has a common limitation: it assumes each document has a single author. While this is sufficient for many applications, in the modern world there are many documents we may wish to attribute which are written collaboratively. With the ubiquitous use of the Internet, content creation platforms such as Wikipedia and GitHub have opened new avenues of collaboration. In this thesis, we examine three authorship attribution problems on two collaborative platforms.