A Computational Approach Based on Syntactic Levels of Language in Authorship Attribution

Abstract

This paper aims to insert a new approach based on syntactic features of the language, which are relating to the essential terms, integrant and accessories of a sentence, such as: subject, predicate and accessories, for the resolution of cases involving the authorship attribution. To this, a database in Portuguese was collected for the experiments. The proposed approach consists of conducting experiments with various fusion methods (sum, mean, median and majority vote), to verify the best performance and method for both authorship verification and identification. To evaluate the approach we used the dependent and independent models. The results generated in authorship verification were between 89-95% accuracy, and 81-90% in the authorship identification through classifiers based on SVM ‑ Support Vector Machines.