Language and task independent text categorization with simple language models

Abstract

We present a simple method for language independent and task independent text categorization learning, based on character-level n-gram language models. Our approach uses simple information theoretic principles and achieves effective performance across a variety of languages and tasks without requiring feature selection or extensive pre-processing. To demonstrate the language and task independence of the proposed technique, we present experimental results on several languages---Greek, English, Chinese and Japanese---in several text categorization problems---language identification, authorship attribution, text genre classification, and topic detection. Our experimental results show that the simple approach achieves state of the art performance in each case.