起因
需要在ES中使用聚合进行统计分析,但是聚合字段值为中文,ES的默认分词器对于中文支持非常不友好:会把完整的中文词语拆分为一系列独立的汉字进行聚合,显然这并不是我的初衷。我们来看个实例:
POST http://192.168.80.133:9200/my_index_name/my_type_name/_search
{
"size": 0,
"query" : {
"range" : {
"time": {
"gte": 1513778040000,
"lte": 1513848720000
}
}
},
"aggs": {
"keywords": {
"terms": {"field": "keywords"},
"aggs": {
"emotions": {
"terms": {"field": "emotion"}
}
}
}
}
}
输出结果:
{
"took": 22,
"timed_out": false,
"_shards": {
"total": 5,
"successful": 5,
"failed": 0
},
"hits": {
"total": 32,
"max_score": 0.0,
"hits": []
},
"aggregations": {
"keywords": {
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0,
"buckets": [
{
"key": "力", # 完整的词被拆分为独立的汉字
"doc_count": 2,
"emotions": {
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0,
"buckets": [
{
"key": -1,
"doc_count": 1
},
{
"key": 0,
"doc_count": 1
}
]
}
},
{
"key": "动",
"doc_count": 2,
"emotions": {
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0,
"buckets": [
{
"key": -1,
"doc_count": 1
},
{
"key": 0,
"doc_count": 1
}
]
}
}
]
}
}
}
既然ES的默认分词器对于中文支持非常不友好,那么有没有可以支持中文的分词器呢?如果有,该如何使用呢?
第一个问题,万能的谷歌告诉了我结果,已经有了支持中文的分词器,而且是开源实现:IK Analysis for Elasticsearch,详见:https://github.com/medcl/elasticsearch-analysis-ik。
秉着“拿来主义”不重复造轮子的指导思想,直接先拿过来使用一下,看看效果怎么样。那么,如何使用IK分词器呢?其实这是一个ES插件,直接安装并对ES进行相应的配置即可。
安装IK分词器
我的ES版本为2.4.1,需要下载的IK版本为:1.10.1(注意:必须下载与ES版本对应的IK,否则不能使用)。
1.下载,编译IK
wget https://github.com/medcl/elasticsearch-analysis-ik/releases/download/v1.10.1/elasticsearch-analysis-ik-1.10.1.zip
unzip elasticsearch-analysis-ik-1.10.1.zip
cd elasticsearch-analysis-ik-1.10.1
mvn clean package
在elasticsearch-analysis-ik-1.10.1\\target\\releases目录下生成打包文件:elasticsearch-analysis-ik-1.10.1.zip。
2.在ES中安装IK插件
将上述打包好的IK插件:elasticsearch-analysis-ik-1.10.1.zip拷贝到ES/plugins目录下,执行解压。
unzip elasticsearch-analysis-ik-1.10.1.zip
rm -rf elasticsearch-analysis-ik-1.10.1.zip # 解压完之后一定要删除这个zip包,否则在启动ES时报错
重启ES。
使用IK分词器
安装IK分词器完毕之后,就可以在ES使用了。
第一步:新建index
PUT http://192.168.80.133:9200/my_index_name
第二步:给将来要使用的doc字段添加mapping
在这里我在ES中存储的doc格式如下:
{
"nagtive_kw": []
"is_all": false,
"emotion": 0,
"focuce": false,
"keywords": ["动力","外观","油耗"], // 在keywords字段上进行聚合分析
"source": "汽车之家",
"time": -1,
"machine_emotion": 0,
"title": "no title",
"spider": "qczj_index",
"content": {},
"url": "http://xxx",
"brand": "宝马",
"series": "宝马1系",
"model": "2017款"
}
需要在keywords字段上进行聚合分析,所以给keywords字段添加mapping设置:
POST http://192.168.80.133:9200/my_index_name/my_type_name/_mapping
{
"properties": {
"keywords": { # 设置keywords字段使用ik分词器
"type": "string",
"store": "no",
"analyzer": "ik_smart",
"search_analyzer": "ik_smart",
"boost": 8
}
}
}
注意: 在设置mapping时有一个小插曲,我根据IK的官网设置“keywords”的type为“text”时报错:
POST http://192.168.80.133:9200/my_index_name/my_type_name/_mapping
{
"properties": {
"keywords": {
"type": "text", # text类型在2.4.1版本中不支持
"store": "no",
"analyzer": "ik_smart",
"search_analyzer": "ik_smart",
"boost": 8
}
}
}
报错:
{
"error": {
"root_cause": [
{
"type": "mapper_parsing_exception",
"reason": "No handler for type [text] declared on field [keywords]"
}
],
"type": "mapper_parsing_exception",
"reason": "No handler for type [text] declared on field [keywords]"
},
"status": 400
}
这是因为我使用的ES版本比较低:2.4.1,而text
类型是ES5.0之后才添加的类型,所以不支持。在ES2.4.1版本中需要使用string
类型。
第三步:添加doc对象
POST http://192.168.80.133:9200/my_index_name/my_type_name/
{
"nagtive_kw": ["动力","外观","油耗"]
"is_all": false,
"emotion": 0,
"focuce": false,
"keywords": ["动力","外观","油耗"], // 在keywords字段上进行聚合分析
"source": "汽车之家",
"time": -1,
"machine_emotion": 0,
"title": "从动次打次吃大餐",
"spider": "qczj_index",
"content": {},
"url": "http://xxx",
"brand": "宝马",
"series": "宝马1系",
"model": "2017款"
}
第四步:聚合分析
POST http://192.168.80.133:9200/my_index_name/my_type_name/_search
{
"size": 0,
"query" : {
"range" : {
"time": {
"gte": 1513778040000,
"lte": 1513848720000
}
}
},
"aggs": {
"keywords": {
"terms": {"field": "keywords"},
"aggs": {
"emotions": {
"terms": {"field": "emotion"}
}
}
}
}
}
输出结果:
{
"took": 22,
"timed_out": false,
"_shards": {
"total": 5,
"successful": 5,
"failed": 0
},
"hits": {
"total": 32,
"max_score": 0.0,
"hits": []
},
"aggregations": {
"keywords": {
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0,
"buckets": [
{
"key": "动力", # 完整的词没有被拆分为独立的汉字
"doc_count": 2,
"emotions": {
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0,
"buckets": [
{
"key": -1,
"doc_count": 1
},
{
"key": 0,
"doc_count": 1
}
]
}
}
]
}
}
}
【参考】
http://www.cnblogs.com/xing901022/p/5910139.html 如何在Elasticsearch中安装中文分词器(IK+pinyin)
https://elasticsearch.cn/question/47 关于聚合(aggs)的问题
https://github.com/medcl/elasticsearch-analysis-ik/issues/276 create map时出现No handler for type [text] declared on field [content] #276
http://blog.csdn.net/guo_jia_liang/article/details/52980716 Elasticsearch2.4学习(三)------Elasticsearch2.4插件安装详解