Using domain knowledge and bilingual resources for addressing community question answering for Arabic

Abstract

This paper presents a description of the approach of the UPC-USMBA team for addressing Community Question Answering, for the Arabic language on the medical domain. Our approach for addressing the task is based on combining the use of original Arabic texts with English translations over which supervised Machine Learning techniques are applied. Our system perform on four steps: A preliminary step, aiming to collect domain resources, a learning step, for getting two models, one over Arabic texts and the other on English texts, a classification step, for applying them to the test datasets, and, finally a combination step over the results of the two classifiers.