For these cases specifically, we use machine learning. There are many cases.
j 下一段next speechk 上一段previous speech
