| In this paper, we seek to describe, test, evaluate, and compare methods of open class attribution that utilize multiple unique closed class attributions in a voting framework. By applying statistical techniques to the proportion of closed class attributions indicating individual candidate authors, we seek to determine if the author is present in a set of suspected authors or not. The final answer to an open class attribution problem is either one of the authors in the set of candidate authors or "None of the above." We test nine different methods of open class attribution grouped into three distinct voting paradigms. We find that the most effective method is a voting method in which each closed class attribution votes equally for its top two most likely authors. Accuracies in this method are statistically better than chance and, in total, are the best out of all nine methods. |
