一、拼音分词的应用

拼音分词在日常生活中其实很常见，也许你每天都在用。打开淘宝看一看吧,输入拼音”zhonghua”,下面会有包含”zhonghua”对应的中文”中华”的商品的提示：

拼音分词是根据输入的拼音提示对应的中文，通过拼音分词提升搜索体验、加快搜索速度。下面介绍如何在Elasticsearch 5.1.1中配置和实现pinyin+iK分词。

二、IK分词器下载与安装

关于IK分词器的介绍不再多少，一言以蔽之，IK分词是目前使用非常广泛分词效果比较好的中文分词器。做ES开发的，中文分词十有八九使用的都是IK分词器。

下载地址:https://github.com/medcl/elasticsearch-analysis-ik
配置之前关闭elasticsearch，配置完成以后再重启。
IK的版本要和当前ES的版本一致，README中有说明。我使用的是ES是5.1.1，IK的版本为5.1.1(你也许会奇怪为什么IK上一个版本是1.X,下一个版本一下升到5.X?是因为Elastic官方为了统一版本号，之前es的版本是2.x,logstash的版本是2.x,同时Kibana的版本是4.x，ik的版本是1.x，这样版本很混乱。5.0之后，统一版本号，这样你使用5.1.1的es，其它软件的版本也使用5.1.1就好了)。

下载之后进入到elasticsearch-analysis-pinyin-master目录，mvn打包(没有安装maven的自行安装)，运行命令：

    mvn package

打包成功以后，会生成一个target文件夹，在elasticsearch-analysis-ik-master/target/releases目录下，找到elasticsearch-analysis-ik-5.1.1.zip，这就是我们需要的安装文件。解压elasticsearch-analysis-ik-5.1.1.zip，得到下面内容：

commons-codec-1.9.jar
commons-logging-1.2.jar
config
elasticsearch-analysis-ik-5.1.1.jar
httpclient-4.5.2.jar
httpcore-4.4.4.jar
plugin-descriptor.properties

然后在elasticsearch-5.1.1/plugins目录下新建一个文件夹ik，把elasticsearch-analysis-ik-5.1.1.zip解压后的文件拷贝到elasticsearch-5.1.1/plugins/ik目录下.截图方便理解。

三、pinyin分词器下载与安装

pinyin分词器的下载地址:
https://github.com/medcl/elasticsearch-analysis-pinyin

安装过程和IK一样，下载、打包、加入ES。这里不在重复上述步骤，给出最后配置截图

四、分词测试

IK和pinyin分词配置完成以后，重启ES。如果重启过程中ES报错，说明安装有错误，没有报错说明配置成功。

4.1 IK分词测试

创建一个索引:

curl -XPUT "http://localhost:9200/index"

测试分词效果:

curl -XPOST "http://localhost:9200/index/_analyze?analyzer=ik_max_word&text=中华人民共和国"

分词结果:

   {"tokens": [{"token": "中华人民共和国","start_offset": 0,"end_offset": 7,"type": "CN_WORD","position": 0}, {"token": "中华人民","start_offset": 0,"end_offset": 4,"type": "CN_WORD","position": 1}, {"token": "中华","start_offset": 0,"end_offset": 2,"type": "CN_WORD","position": 2}, {"token": "华人","start_offset": 1,"end_offset": 3,"type": "CN_WORD","position": 3}, {"token": "人民共和国","start_offset": 2,"end_offset": 7,"type": "CN_WORD","position": 4}, {"token": "人民","start_offset": 2,"end_offset": 4,"type": "CN_WORD","position": 5}, {"token": "共和国","start_offset": 4,"end_offset": 7,"type": "CN_WORD","position": 6}, {"token": "共和","start_offset": 4,"end_offset": 6,"type": "CN_WORD","position": 7}, {"token": "国","start_offset": 6,"end_offset": 7,"type": "CN_CHAR","position": 8}, {"token": "国歌","start_offset": 7,"end_offset": 9,"type": "CN_WORD","position": 9}]
}

使用ik_smart分词:

curl -XPOST "http://localhost:9200/index/_analyze?analyzer=ik_smart&text=中华人民共和国"

分词结果:

{"tokens": [{"token": "中华人民共和国","start_offset": 0,"end_offset": 7,"type": "CN_WORD","position": 0}, {"token": "国歌","start_offset": 7,"end_offset": 9,"type": "CN_WORD","position": 1}]
}

截图方便理解:

4.2拼音分词测试

测试拼音分词:

curl -XPOST "http://localhost:9200/index/_analyze?analyzer=pinyin&text=张学友"

分词结果:

{"tokens": [{"token": "zhang","start_offset": 0,"end_offset": 1,"type": "word","position": 0}, {"token": "xue","start_offset": 1,"end_offset": 2,"type": "word","position": 1}, {"token": "you","start_offset": 2,"end_offset": 3,"type": "word","position": 2}, {"token": "zxy","start_offset": 0,"end_offset": 3,"type": "word","position": 3}]
}

五、IK+pinyin分词配置

5.1创建索引与分析器设置

创建一个索引，并设置index分析器相关属性:

curl -XPUT "http://localhost:9200/medcl/" -d'
{"index": {"analysis": {"analyzer": {"ik_pinyin_analyzer": { "type": "custom", "tokenizer": "ik_smart", "filter": ["my_pinyin", "word_delimiter"] }},"filter": {"my_pinyin": { "type": "pinyin", "first_letter": "prefix", "padding_char": " " }}}}
}'

创建一个type并设置mapping:

curl -XPOST http://localhost:9200/medcl/folks/_mapping -d'
{"folks": {"properties": {"name": {"type": "keyword","fields": {"pinyin": {"type": "text","store": "no","term_vector": "with_positions_offsets","analyzer": "ik_pinyin_analyzer","boost": 10}}}}}
}'

5.2索引测试文档

索引2份测试文档。
文档1:

curl -XPOST http://localhost:9200/medcl/folks/andy -d'{"name":"刘德华"}'

文档2:

curl -XPOST http://localhost:9200/medcl/folks/tina -d'{"name":"中华人民共和国国歌"}'

5.3测试(1)拼音分词

下面四条命命令都可以匹配”刘德华”

curl -XPOST "http://localhost:9200/medcl/folks/_search?q=name.pinyin:liu"curl -XPOST "http://localhost:9200/medcl/folks/_search?q=name.pinyin:de"curl -XPOST "http://localhost:9200/medcl/folks/_search?q=name.pinyin:hua"curl -XPOST "http://localhost:9200/medcl/folks/_search?q=name.pinyin:ldh"

5.4测试(2)IK分词测试

curl -XPOST "http://localhost:9200/medcl/_search?pretty" -d'
{"query": {"match": {"name.pinyin": "国歌"}},"highlight": {"fields": {"name.pinyin": {}}}
}'

返回结果:

{"took" : 2,"timed_out" : false,"_shards" : {"total" : 5,"successful" : 5,"failed" : 0},"hits" : {"total" : 1,"max_score" : 16.698704,"hits" : [{"_index" : "medcl","_type" : "folks","_id" : "tina","_score" : 16.698704,"_source" : {"name" : "中华人民共和国国歌"},"highlight" : {"name.pinyin" : ["<em>中华人民共和国</em><em>国歌</em>"]}}]}
}

说明IK分词器起到了效果。

5.3测试(4)pinyin+ik分词测试：

curl -XPOST "http://localhost:9200/medcl/_search?pretty" -d'
{"query": {"match": {"name.pinyin": "zhonghua"}},"highlight": {"fields": {"name.pinyin": {}}}
}'

返回结果:

{"took" : 3,"timed_out" : false,"_shards" : {"total" : 5,"successful" : 5,"failed" : 0},"hits" : {"total" : 2,"max_score" : 5.9814634,"hits" : [{"_index" : "medcl","_type" : "folks","_id" : "tina","_score" : 5.9814634,"_source" : {"name" : "中华人民共和国国歌"},"highlight" : {"name.pinyin" : ["<em>中华人民共和国</em>国歌"]}},{"_index" : "medcl","_type" : "folks","_id" : "andy","_score" : 2.2534127,"_source" : {"name" : "刘德华"},"highlight" : {"name.pinyin" : ["<em>刘德华</em>"]}}]}
}

截图如下:

使用pinyin分词以后，原始的字段搜索要加上.pinyin后缀，搜索原始字段没有返回结果：

六、参考资料

1. https://github.com/medcl/elasticsearch-analysis-ik
2. https://github.com/medcl/elasticsearch-analysis-pinyin
3. https://my.oschina.net/xiaohui249/blog/214505