Combining Naive Bayes and n-Gram Language Models for Text Classification