使用 Spark 将句子编码为序列模型

Encode sentence as sequence model with Spark

我正在做文本分类,我使用 pyspark.ml.feature.Tokenizer 来标记文本。但是 CountVectorizer 将标记化的单词列表转换为词袋模型,而不是序列模型。

假设我们有以下包含列 id 和文本的 DataFrame:

 id | texts
----|----------
 0  | Array("a", "b", "c")
 1  | Array("a", "b", "b", "c", "a")
each row in texts is a document of type Array[String]. Invoking fit of CountVectorizer produces a CountVectorizerModel with vocabulary (a, b, c). Then the output column “vector” after transformation contains:

 id | texts                           | vector
----|---------------------------------|---------------
 0  | Array("a", "b", "c")            | (3,[0,1,2],[1.0,1.0,1.0])
 1  | Array("a", "b", "b", "c", "a")  | (3,[0,1,2],[2.0,2.0,1.0])

我想要的是(第 1 行)

Array("a", "b", "b", "c", "a")  | [0, 1, 1, 2, 0]

那么我是否可以将自定义函数并行写入 运行 编码?或者除了使用 spark 之外,还有其他可以并行执行的库吗?

您可以使用 StringIndexerexplode:

df = spark_session.createDataFrame([
    Row(id=0, texts=["a", "b", "c"]),
    Row(id=1, texts=["a", "b", "b", "c", "a"])
])

data = df.select("id", explode("texts").alias("texts"))
indexer = StringIndexer(inputCol="texts", outputCol="indexed", stringOrderType="alphabetAsc")
indexer\
    .fit(data)\
    .transform(data)\
    .groupBy("id")\
    .agg(collect_list("texts").alias("texts"), collect_list("indexed").alias("vector"))\
    .show(20, False)

输出:

+---+---------------+-------------------------+
|id |texts          |vector                   |
+---+---------------+-------------------------+
|0  |[a, b, c]      |[0.0, 1.0, 2.0]          |
|1  |[a, b, b, c, a]|[0.0, 1.0, 1.0, 2.0, 0.0]|
+---+---------------+-------------------------+