首发于六哥随笔
ElasticSearch的match和match_phrase查询

ElasticSearch的match和match_phrase查询

问题:

索引中有『第十人民医院』这个字段,使用IK分词结果如下 :

POST http://localhost:9200/development_hospitals/_analyze?pretty&field=hospital.names&analyzer=ik

{
  "tokens": [
    {
      "token": "第十",
      "start_offset": 0,
      "end_offset": 2,
      "type": "CN_WORD",
      "position": 0
    },
    {
      "token": "十人",
      "start_offset": 1,
      "end_offset": 3,
      "type": "CN_WORD",
      "position": 1
    },
    {
      "token": "十",
      "start_offset": 1,
      "end_offset": 2,
      "type": "TYPE_CNUM",
      "position": 2
    },
    {
      "token": "人民医院",
      "start_offset": 2,
      "end_offset": 6,
      "type": "CN_WORD",
      "position": 3
    },
    {
      "token": "人民",
      "start_offset": 2,
      "end_offset": 4,
      "type": "CN_WORD",
      "position": 4
    },
    {
      "token": "人",
      "start_offset": 2,
      "end_offset": 3,
      "type": "COUNT",
      "position": 5
    },
    {
      "token": "民医院",
      "start_offset": 3,
      "end_offset": 6,
      "type": "CN_WORD",
      "position": 6
    },
    {
      "token": "医院",
      "start_offset": 4,
      "end_offset": 6,
      "type": "CN_WORD",
      "position": 7
    }
  ]
}

使用Postman构建match查询:

可以得到结果,但是使用match_phrase查询『第十』却没有任何结果

问题分析:

参考文档 The Definitive Guide [2.x] | Elastic

phrase搜索跟关键字的位置有关, 『第十』采用ik_max_word分词结果如下

{
  "tokens": [
    {
      "token": "第十",
      "start_offset": 0,
      "end_offset": 2,
      "type": "CN_WORD",
      "position": 0
    },
    {
      "token": "十",
      "start_offset": 1,
      "end_offset": 2,
      "type": "TYPE_CNUM",
      "position": 1
    }
  ]
}

虽然『第十』和『十』都可以命中,但是match_phrase的特点是分词后的相对位置也必须要精准匹配,『第十人民医院』采用id_max_word分词后,『第十』和『十』之间有一个『十人』,所以无法命中。

解决方案:

采用ik_smart分词可以避免这样的问题,对『第十人民医院』和『第十』采用ik_smart分词的结果分别是:

{
  "tokens": [
    {
      "token": "第十",
      "start_offset": 0,
      "end_offset": 2,
      "type": "CN_WORD",
      "position": 0
    },
    {
      "token": "人民医院",
      "start_offset": 2,
      "end_offset": 6,
      "type": "CN_WORD",
      "position": 1
    }
  ]
}
{
  "tokens": [
    {
      "token": "第十",
      "start_offset": 0,
      "end_offset": 2,
      "type": "CN_WORD",
      "position": 0
    }
  ]
}

稳稳命中

最佳实践:

采用match_phrase匹配,结果会非常严格,但是也会漏掉相关的结果,个人觉得混合两种方式进行bool查询比较好,并且对match_phrase匹配采用boost加权,比如对name进行2种分词并索引,ik_smart分词采用match_phrase匹配,ik_max_word分词采用match匹配,如:

{
  "query": {
    "bool": {
      "should": [
          {"match_phrase": {"name1": {"query": "第十", "boost": 2}}},
          {"match": {"name2": "第十"}}
      ]
    }
  },
  explain: true
}
编辑于 2017-03-24 18:03