ITALICA at PAN 2013: An Ensemble Learning Approach to Author Profiling Notebook for PAN at CLEF 2013

Abstract

This notebook discusses the approach to the Author Profiling task developed by the Italica group for PAN 2013. This system implements two different sets of classifiers which are combined later in order to build a final classifier that takes into account the decisions of the previous ones. The initial classifiers are focused on vector space representations of the documents as a bag of words and n-grams of POS tags and also on a set of stylistic features of the texts. The final classifier consists of a stacking schema that combines the other ones. This approach has obtained better results for the Spanish dataset than for the English dataset, probably due to the use of more detailed POS tagset in the former