文本特征生成预测 (DocCountVectorizerPredictBatchOp)

Java 类名：com.alibaba.alink.operator.batch.nlp.DocCountVectorizerPredictBatchOp

Python 类名：DocCountVectorizerPredictBatchOp

功能介绍

根据文本中词语的特征信息，将每条文本转化为稀疏向量。

使用方式

该组件是预测组件，需要配合训练组件 DocCountVectorizerTrainBatchOp 使用。

参数说明

名称	中文名称	描述	类型	是否必须？	取值范围	默认值
selectedCol	选中的列名	计算列对应的列名	String	✓	所选列类型为 [STRING]
modelFilePath	模型的文件路径	模型的文件路径	String			null
outputCol	输出结果列	输出结果列列名，可选，默认null	String			null
reservedCols	算法保留列名	算法保留列	String[]			null
numThreads	组件多线程线程个数	组件多线程线程个数	Integer			1

代码示例

Python 代码

df = pd.DataFrame([
    [0, u'二手旧书:医学电磁成像'],
    [1, u'二手美国文学选读（ 下册 ）李宜燮南开大学出版社 9787310003969'],
    [2, u'二手正版图解象棋入门/谢恩思主编/华龄出版社'],
    [3, u'二手中国糖尿病文献索引'],
    [4, u'二手郁达夫文集（ 国内版 ）全十二册馆藏书']
])

inOp1 = BatchOperator.fromDataframe(df, schemaStr='id int, text string')

segment = SegmentBatchOp().setSelectedCol("text").linkFrom(inOp1)
train = DocCountVectorizerTrainBatchOp().setSelectedCol("text").linkFrom(segment)
predictBatch = DocCountVectorizerPredictBatchOp().setSelectedCol("text").linkFrom(train, segment)
train.lazyPrint(-1)
predictBatch.print()

Java 代码

import org.apache.flink.types.Row;

import com.alibaba.alink.operator.batch.BatchOperator;
import com.alibaba.alink.operator.batch.nlp.DocCountVectorizerPredictBatchOp;
import com.alibaba.alink.operator.batch.nlp.DocCountVectorizerTrainBatchOp;
import com.alibaba.alink.operator.batch.nlp.SegmentBatchOp;
import com.alibaba.alink.operator.batch.source.MemSourceBatchOp;
import org.junit.Test;

import java.util.Arrays;
import java.util.List;

public class DocCountVectorizerPredictBatchOpTest {
	@Test
	public void testDocCountVectorizerPredictBatchOp() throws Exception {
		List <Row> df = Arrays.asList(
			Row.of(0, "二手旧书:医学电磁成像"),
			Row.of(1, "二手美国文学选读（ 下册 ）李宜燮南开大学出版社 9787310003969"),
			Row.of(2, "二手正版图解象棋入门/谢恩思主编/华龄出版社"),
			Row.of(3, "二手中国糖尿病文献索引"),
			Row.of(4, "二手郁达夫文集（ 国内版 ）全十二册馆藏书")
		);
		BatchOperator <?> inOp1 = new MemSourceBatchOp(df, "id int, text string");
		BatchOperator <?> segment = new SegmentBatchOp().setSelectedCol("text").linkFrom(inOp1);
		BatchOperator <?> train = new DocCountVectorizerTrainBatchOp().setSelectedCol("text").linkFrom(segment);
		BatchOperator <?> predictBatch = new DocCountVectorizerPredictBatchOp().setSelectedCol("text").linkFrom(train,
			segment);
		train.lazyPrint(-1);
		predictBatch.print();
	}
}

运行结果

模型数据

model_id	model_info
0	{“minTF”:“1.0”,“featureType”:“"WORD_COUNT"”}
1048576	{“f0”:“）”,“f1”:0.6931471805599453,“f2”:0}
2097152	{“f0”:“（”,“f1”:0.6931471805599453,“f2”:1}
3145728	{“f0”:“馆藏”,“f1”:1.0986122886681098,“f2”:2}
4194304	{“f0”:“郁达夫”,“f1”:1.0986122886681098,“f2”:3}
5242880	{“f0”:“选读”,“f1”:1.0986122886681098,“f2”:4}
6291456	{“f0”:“象棋”,“f1”:1.0986122886681098,“f2”:5}
7340032	{“f0”:“谢恩”,“f1”:1.0986122886681098,“f2”:6}
8388608	{“f0”:“美国”,“f1”:1.0986122886681098,“f2”:7}
9437184	{“f0”:“索引”,“f1”:1.0986122886681098,“f2”:8}
10485760	{“f0”:“糖尿病”,“f1”:1.0986122886681098,“f2”:9}
11534336	{“f0”:“电磁”,“f1”:1.0986122886681098,“f2”:10}
12582912	{“f0”:“版”,“f1”:1.0986122886681098,“f2”:11}
13631488	{“f0”:“正版”,“f1”:1.0986122886681098,“f2”:12}
14680064	{“f0”:“李宜燮”,“f1”:1.0986122886681098,“f2”:13}
15728640	{“f0”:“旧书”,“f1”:1.0986122886681098,“f2”:14}
16777216	{“f0”:“文集”,“f1”:1.0986122886681098,“f2”:15}
17825792	{“f0”:“文献”,“f1”:1.0986122886681098,“f2”:16}
18874368	{“f0”:“文学”,“f1”:1.0986122886681098,“f2”:17}
19922944	{“f0”:“成像”,“f1”:1.0986122886681098,“f2”:18}
20971520	{“f0”:“思”,“f1”:1.0986122886681098,“f2”:19}
22020096	{“f0”:“图解”,“f1”:1.0986122886681098,“f2”:20}
23068672	{“f0”:“国内”,“f1”:1.0986122886681098,“f2”:21}
24117248	{“f0”:“南开大学”,“f1”:1.0986122886681098,“f2”:22}
25165824	{“f0”:“华龄”,“f1”:1.0986122886681098,“f2”:23}
26214400	{“f0”:“十二册”,“f1”:1.0986122886681098,“f2”:24}
27262976	{“f0”:“医学”,“f1”:1.0986122886681098,“f2”:25}
28311552	{“f0”:“出版社”,“f1”:0.6931471805599453,“f2”:26}
29360128	{“f0”:“全”,“f1”:1.0986122886681098,“f2”:27}
30408704	{“f0”:“入门”,“f1”:1.0986122886681098,“f2”:28}
31457280	{“f0”:“二手”,“f1”:0.0,“f2”:29}
32505856	{“f0”:“书”,“f1”:1.0986122886681098,“f2”:30}
33554432	{“f0”:“主编”,“f1”:1.0986122886681098,“f2”:31}
34603008	{“f0”:“中国”,“f1”:1.0986122886681098,“f2”:32}
35651584	{“f0”:“下册”,“f1”:1.0986122886681098,“f2”:33}
36700160	{“f0”:“:”,“f1”:1.0986122886681098,“f2”:34}
37748736	{“f0”:“9787310003969”,“f1”:1.0986122886681098,“f2”:35}
38797312	{“f0”:“/”,“f1”:1.0986122886681098,“f2”:36}

批预测结果

id	text
0	$37$10:1.0 14:1.0 18:1.0 25:1.0 29:1.0 34:1.0
1	$37$0:1.0 1:1.0 4:1.0 7:1.0 13:1.0 17:1.0 22:1.0 26:1.0 29:1.0 33:1.0 35:1.0
2	$37$5:1.0 6:1.0 12:1.0 19:1.0 20:1.0 23:1.0 26:1.0 28:1.0 29:1.0 31:1.0 36:2.0
3	$37$8:1.0 9:1.0 16:1.0 29:1.0 32:1.0
4	$37$0:1.0 1:1.0 2:1.0 3:1.0 11:1.0 15:1.0 21:1.0 24:1.0 27:1.0 29:1.0 30:1.0