Extracting key-substring-group features for text classification