Complex linguistic features for text classification: a comprehensive study