Graph-based and Lexical-Syntactic Approaches for the Authorship Attribution Task

In this paper we present two different approaches for tackling the authorship attribution task. The first approach uses a set of phrase-level lexicalsyntactic features, whereas the second approach considers a graph-based text representation together with a data mining technique for discovering authorship patterns which may be further used for attributing the authorship of an anonymous document. In both cases we employed a support vector machine classifier in order to determine the target class. The features extracted by means of the graph-based approach allowed it to obtain a better performance than the other approach.