Is it possible to use this method to segment corpus data without any segmentation to sentences and words?