Application of DRSA-ANN Classifier in Computational Stylistics

摘要

Computational stylistics or stylometry deals with characteristics of writing styles. It assumes that each author expresses themselves in such an individual way that a writing style can be uniquely defined and described by some quantifiable measures. With help of contemporary computers the stylometric tasks of author characterisation, comparison, and attribution can be implemented using either some statistic-oriented approaches or methodologies from artificial intelligence domain. The paper presents results of research on an application of a hybrid classifier, combining Dominance-based Rough Set Approach and Artificial Neural Networks, within the task of authorship attribution for literary texts. The performance of the classifier is observed while exploiting an analysis of characteristic features basing on the cardinalities of relative reducts found within rough set processing.