Classification of texts potentially containing a complex and specific terminology requires the use of learning methods that do not rely on extensive feature engineering. In this work we use prediction by partial matching ...