考虑使用不同长度的短语训练的RNN来预测下一个字母，给定一个字符串。
假设我有以下短语作为我的数据集：
I really want to go to work
the book has arrived
I will buy the red car
输入序列为：
I really want to go to wor
the book has arrive
I will buy the red ca
标签是：
really want to go to work
he book has arrived
will buy the red car
可以用tensorflow完成吗？切片函数似乎只适用于相同长度的句子。