创建PySpark ML定制变压器 [英] Create a custom Transformer in PySpark ML
问题描述
我是新来的星火他们(PySpark)SQL DataFrames和ML。
我怎样才能建立一个服装标记生成器,这对于例如去除停止词,并使用一些图书馆从的 NLTK ?我可以延长默认的?
感谢。
我可以扩展默认的?
块引用>不是真的。默认
标记生成器
是pyspark.ml.wrapper.JavaTransformer
的一个子类,并像其他transfromers和估计从pyspark.ml.feature
,代表实际加工到斯卡拉对应。由于要使用Python,你应该延长pyspark.ml.pipeline.Transformer
直接。进口NLTK
从pyspark.ml.pipeline进口变压器
从pyspark.ml.param.shared进口HasInputCol,HasOutputCol,参数
从pyspark.ml.util进口keyword_only从pyspark.sql.functions进口UDF
从pyspark.sql.types输入数组类型,StringType类NLTKWordPunctTokenizer(变压器,HasInputCol,HasOutputCol): @keyword_only
高清__init __(自我,inputCol =无,outputCol =无禁用词=无):
超(NLTKWordPunctTokenizer,个体经营).__的init __()
self.stopwords =参数(自我,禁用词,)
self._setDefault(停用词=集())
kwargs =自.__初始化__._ input_kwargs
self.setParams(** kwargs) @keyword_only
高清setParams(个体经营,inputCol =无,outputCol =无禁用词=无):
kwargs = self.setParams._input_kwargs
返回self._set(** kwargs) 高清setStopwords(个体经营,价值):
self._paramMap [self.stopwords] =价值
自回归 高清getStopwords(个体经营):
返回self.getOrDefault(self.stopwords) 高清_transform(个体经营,数据集):
停用词= self.getStopwords() DEF F(S):
令牌= nltk.tokenize.wordpunct_tokenize(多个)
返回[在令牌吨t如此t.lower()不是禁用词] T =数组类型(StringType())
out_col = self.getOutputCol()
in_col =数据集[self.getInputCol()]
返回dataset.withColumn(out_col,UDF(F,T)(in_col))用法示例(从 ML数据 - 功能):
sentenceDataFrame = sqlContext.createDataFrame([
(0,我听说星火),
(0,我希望能Java的使用情况班),
(1,逻辑回归模型整齐)
],[标签,句])标记生成器= NLTKWordPunctTokenizer(
inputCol =句,outputCol =字,
停用词=集(nltk.corpus.stopwords.words('英语')))tokenizer.transform(sentenceDataFrame).show()I am new to Spark SQL DataFrames and ML on them (PySpark). How can I create a costume tokenizer, which for example removes stop words and uses some libraries from nltk? Can I extend the default one?
Thanks.
解决方案Can I extend the default one?
Not really. Default
Tokenizer
is a subclass ofpyspark.ml.wrapper.JavaTransformer
and, same as other transfromers and estimators frompyspark.ml.feature
, delegates actual processing to its Scala counterpart. Since you want to use Python you should extendpyspark.ml.pipeline.Transformer
directly.import nltk from pyspark.ml.pipeline import Transformer from pyspark.ml.param.shared import HasInputCol, HasOutputCol, Param from pyspark.ml.util import keyword_only from pyspark.sql.functions import udf from pyspark.sql.types import ArrayType, StringType class NLTKWordPunctTokenizer(Transformer, HasInputCol, HasOutputCol): @keyword_only def __init__(self, inputCol=None, outputCol=None, stopwords=None): super(NLTKWordPunctTokenizer, self).__init__() self.stopwords = Param(self, "stopwords", "") self._setDefault(stopwords=set()) kwargs = self.__init__._input_kwargs self.setParams(**kwargs) @keyword_only def setParams(self, inputCol=None, outputCol=None, stopwords=None): kwargs = self.setParams._input_kwargs return self._set(**kwargs) def setStopwords(self, value): self._paramMap[self.stopwords] = value return self def getStopwords(self): return self.getOrDefault(self.stopwords) def _transform(self, dataset): stopwords = self.getStopwords() def f(s): tokens = nltk.tokenize.wordpunct_tokenize(s) return [t for t in tokens if t.lower() not in stopwords] t = ArrayType(StringType()) out_col = self.getOutputCol() in_col = dataset[self.getInputCol()] return dataset.withColumn(out_col, udf(f, t)(in_col))
Example usage (data from ML - Features):
sentenceDataFrame = sqlContext.createDataFrame([ (0, "Hi I heard about Spark"), (0, "I wish Java could use case classes"), (1, "Logistic regression models are neat") ], ["label", "sentence"]) tokenizer = NLTKWordPunctTokenizer( inputCol="sentence", outputCol="words", stopwords=set(nltk.corpus.stopwords.words('english'))) tokenizer.transform(sentenceDataFrame).show()
这篇关于创建PySpark ML定制变压器的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!