apache spark提供了tf-idf算法：https://spark.apache.org/docs/latest/ml-features.html#tf-以色列国防军
运行该示例时，它会添加“rawfeatures”和“features”列，并输出以下Dataframe：

|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| l | sentence    | words       | rawFeatures                                             | features   
|---|----------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| 0 | Hi...       | ["hi", ...] | [0, 32, [1, 12, 16, 22, 28], [1, 1, 1, 1, 1]]           | [0, 32, [1, 12, 16, 22, 28, [0.69, 0.69, 0.29, 0.29, 0.29]]                    |
|---|----------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| 0 | I wish...   | [...]  | [0, 32, [11, 15, 16, 22, 28, 29, 31], [1, 1, 1, 1, 1, 1, 1]] | [0, 32, [11, 15, 16, 22, 28, 29, 31], [1, 1, 1, 1, 1, 1, 1]]                   |
|---|----------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| 1 | Logistic... | [...] | [0, 32, [3, 4, 15, 27, 30], [1, 1, 1, 1, 1]]                  | [0, 32, [3, 4, 15, 27, 30], [0.69, 0.69, 0.29, 0.69, 0.69]]                    |
|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------|

我有两个问题：
“rawfeatures”和“features”列是什么？它们中的数组如何与tf idf相关？
（假设“features”列的最后一个元素是tf idf）如何将这个Dataframe转换成这样？

|-----------------------|
| word | label | TF-IDF |
|-----------------------|

本质上，我想要一个每个字有多行的Dataframe，它显示的标签，以及它的tf-idf。
提前感谢：）