-
Notifications
You must be signed in to change notification settings - Fork 820
/
TextSHAP.scala
88 lines (72 loc) · 3 KB
/
TextSHAP.scala
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
// Copyright (C) Microsoft Corporation. All rights reserved.
// Licensed under the MIT License. See LICENSE in project root for information.
package com.microsoft.azure.synapse.ml.explainers
import com.microsoft.azure.synapse.ml.core.schema.DatasetExtensions
import com.microsoft.azure.synapse.ml.logging.FeatureNames
import org.apache.spark.injections.UDFUtils
import org.apache.spark.ml.ComplexParamsReadable
import org.apache.spark.ml.linalg.SQLDataTypes.VectorType
import org.apache.spark.ml.param.shared.HasInputCol
import org.apache.spark.ml.util.Identifiable
import org.apache.spark.sql.DataFrame
import org.apache.spark.sql.functions.{col, explode}
import org.apache.spark.sql.types._
trait TextSHAPParams extends KernelSHAPParams with HasInputCol with HasTokensCol {
self: TextSHAP =>
def setInputCol(value: String): this.type = this.set(inputCol, value)
setDefault(tokensCol -> "tokens")
}
class TextSHAP(override val uid: String)
extends KernelSHAPBase(uid)
with TextSHAPParams
with TextExplainer {
logClass(FeatureNames.Explainers)
def this() = {
this(Identifiable.randomUID("TextSHAP"))
}
override protected def createSamples(df: DataFrame,
idCol: String,
coalitionCol: String,
weightCol: String,
targetClassesCol: String): DataFrame = {
val numSamplesOpt = this.getNumSamplesOpt
val infWeightVal = this.getInfWeight
val samplesUdf = UDFUtils.oldUdf(
{
(tokens: Seq[String]) =>
val effectiveNumSamples = KernelSHAPBase.getEffectiveNumSamples(numSamplesOpt, tokens.size)
val sampler = new KernelSHAPTextSampler(tokens, effectiveNumSamples, infWeightVal)
(1 to effectiveNumSamples).map {
_ =>
val (sampleTokens, features, weight) = sampler.sample
val sampleText = sampleTokens.mkString(" ")
(sampleText, features, weight)
}
},
getSampleSchema(StringType)
)
val samplesCol = DatasetExtensions.findUnusedColumnName("samples", df)
df.withColumn(samplesCol, explode(samplesUdf(col(getTokensCol))))
.select(
col(idCol),
col(samplesCol).getField(sampleField).alias(getInputCol),
col(samplesCol).getField(coalitionField).alias(coalitionCol),
col(samplesCol).getField(weightField).alias(weightCol),
col(targetClassesCol)
)
}
override protected def validateSchema(schema: StructType): Unit = {
super.validateSchema(schema)
require(
schema(getInputCol).dataType == StringType,
s"Field $getInputCol is expected to be string type, but got ${schema(getInputCol).dataType} instead."
)
}
override def transformSchema(schema: StructType): StructType = {
this.validateSchema(schema)
schema
.add(getTokensCol, ArrayType(StringType))
.add(getOutputCol, VectorType)
}
}
object TextSHAP extends ComplexParamsReadable[TextSHAP]