I'm assuming the opposite: bag-of-words is the go-to baseline, but as another commenter pointed out, just separating instances based on vocabulary is not sufficient in today's sophisticated text classification problems.