Author Gender Prediction in an Email Stream Using Neural Networks

Abstract

" In the beginning was the Word, and the Word was with God, and the Word was God " . Thus, John 1:1 1 begins his contribution to the Holy Bible (one of the most-distributed book in the world with hundreds of millions of copies 2), the importance of the word lies in the essence of human beings. The discursive style reflects the profile of the author, who decides, often unconsciously, about how to choose and combine words. This provides valuable information about the personality of the author. In this paper we present our approach to identify age and gender of authors based on their use of language. We propose a representation based on stylistic features and obtain encouraging results with a SVM-based approach on the PAN-AP-13 3 dataset.