在之前的我的一篇文章“Elasticsearch:language ingest processor”。在那篇文章中,我们使用了一个社区的插件来完成这个工作。 在Elastic Stack 7.6的发布中,我们很欣喜地告诉大家:在Elastic Stack中,我们集成了一个语言识别的processor。
动手实践
首先,我们需要安装最新的Elastic Stack 7.6的发行版。我们在Kibana中打入如下的命令:
PUT _ingest/pipeline/my-pipeline-id
{
"description": "describe pipeline",
"processors": [
{
"inference": {
"model_id": "lang_ident_model_1",
"inference_config": {
"classification": {
"num_top_classes": 1
}
},
"field_mappings": {}
}
}
]
}
在上面我们创建了一个叫做my-pipeline-id的pipeline。那么当我们把文档导入到Elasticsearch中时,我们可以引用这个pipeline:
POST test/_doc/1?pipeline=my-pipeline-id
{
"text": "我爱北京天安门"
}
在上面我们把 “我爱北京天安门”写入到Elasticsearch中。我们可以通过如下的命令来检查写入的信息:
GET test/_doc/1/
我们可以看到如下的一个响应:
{
"_index" : "test",
"_type" : "_doc",
"_id" : "1",
"_version" : 2,
"_seq_no" : 1,
"_primary_term" : 1,
"found" : true,
"_source" : {
"text" : "我爱北京天安门",
"ml" : {
"inference" : {
"top_classes" : [
{
"class_name" : "zh",
"class_probability" : 0.9999770918986735,
"class_score" : 0.9999770918986735
}
],
"predicted_value" : "zh",
"model_id" : "lang_ident_model_1"
}
}
}
}
在上面,我们可以看到有一个叫做ml.inference.predicated_value的字段,它显示的语言是zh,也就是中文。
同样我们可以对其它的语言来做一个实验,比如日文:
POST test/_doc/2?pipeline=my-pipeline-id
{
"text": "これはフリーテキストです。"
}
那么我们通过如下的命令来检查输入的文档:
GET test/_doc/2/
上面的返回的结果是:
{
"_index" : "test",
"_type" : "_doc",
"_id" : "2",
"_version" : 2,
"_seq_no" : 3,
"_primary_term" : 1,
"found" : true,
"_source" : {
"text" : "これはフリーテキストです。",
"ml" : {
"inference" : {
"top_classes" : [
{
"class_name" : "ja",
"class_probability" : 0.999997979528696,
"class_score" : 0.999997979528696
}
],
"predicted_value" : "ja",
"model_id" : "lang_ident_model_1"
}
}
}
}
在上面的ml.inference.predicted_value显示的是“ja”,也就是日文。
如果我们想使用pipeline的_simulate来测试的话,那么我们可以使用如下的命令:
POST _ingest/pipeline/_simulate
{
"pipeline": {
"processors": [
{
"inference": {
"model_id": "lang_ident_model_1",
"inference_config": {
"classification": {
"num_top_classes": 1
}
},
"field_mappings": {
}
}
}
]
},
"docs": [
{"_source": {"text": "this is some free text."}},
{"_source": {"text": "c'est du texte libre."}},
{"_source": {"text": "Dies ist ein Freitext."}},
{"_source": {"text": "これはフリーテキストです。"}},
{"_source": {"text": "یہ کچھ مفت متن ہے۔"}},
{"_source": {"text": "我爱北京天安门"}}
]
}
显示的结果是:
{
"docs" : [
{
"doc" : {
"_index" : "_index",
"_type" : "_doc",
"_id" : "_id",
"_source" : {
"text" : "this is some free text.",
"ml" : {
"inference" : {
"top_classes" : [
{
"class_name" : "en",
"class_probability" : 0.9999832919481261,
"class_score" : 0.9999832919481261
}
],
"predicted_value" : "en",
"model_id" : "lang_ident_model_1"
}
}
},
"_ingest" : {
"timestamp" : "2020-03-04T05:51:26.380985Z"
}
}
},
{
"doc" : {
"_index" : "_index",
"_type" : "_doc",
"_id" : "_id",
"_source" : {
"text" : "c'est du texte libre.",
"ml" : {
"inference" : {
"top_classes" : [
{
"class_name" : "fr",
"class_probability" : 0.996991483545992,
"class_score" : 0.996991483545992
}
],
"predicted_value" : "fr",
"model_id" : "lang_ident_model_1"
}
}
},
"_ingest" : {
"timestamp" : "2020-03-04T05:51:26.380988Z"
}
}
},
{
"doc" : {
"_index" : "_index",
"_type" : "_doc",
"_id" : "_id",
"_source" : {
"text" : "Dies ist ein Freitext.",
"ml" : {
"inference" : {
"top_classes" : [
{
"class_name" : "de",
"class_probability" : 0.9999995801281829,
"class_score" : 0.9999995801281829
}
],
"predicted_value" : "de",
"model_id" : "lang_ident_model_1"
}
}
},
"_ingest" : {
"timestamp" : "2020-03-04T05:51:26.38099Z"
}
}
},
{
"doc" : {
"_index" : "_index",
"_type" : "_doc",
"_id" : "_id",
"_source" : {
"text" : "これはフリーテキストです。",
"ml" : {
"inference" : {
"top_classes" : [
{
"class_name" : "ja",
"class_probability" : 0.999997979528696,
"class_score" : 0.999997979528696
}
],
"predicted_value" : "ja",
"model_id" : "lang_ident_model_1"
}
}
},
"_ingest" : {
"timestamp" : "2020-03-04T05:51:26.380992Z"
}
}
},
{
"doc" : {
"_index" : "_index",
"_type" : "_doc",
"_id" : "_id",
"_source" : {
"text" : "یہ کچھ مفت متن ہے۔",
"ml" : {
"inference" : {
"top_classes" : [
{
"class_name" : "ur",
"class_probability" : 0.9999762917090316,
"class_score" : 0.9999762917090316
}
],
"predicted_value" : "ur",
"model_id" : "lang_ident_model_1"
}
}
},
"_ingest" : {
"timestamp" : "2020-03-04T05:51:26.380994Z"
}
}
},
{
"doc" : {
"_index" : "_index",
"_type" : "_doc",
"_id" : "_id",
"_source" : {
"text" : "我爱北京天安门",
"ml" : {
"inference" : {
"top_classes" : [
{
"class_name" : "zh",
"class_probability" : 0.9999770918986735,
"class_score" : 0.9999770918986735
}
],
"predicted_value" : "zh",
"model_id" : "lang_ident_model_1"
}
}
},
"_ingest" : {
"timestamp" : "2020-03-04T05:51:26.380995Z"
}
}
}
]
}