Elasticsearch搜索推荐

本文基于Elasticsearch7.x版本.

什么是搜索推荐?

当我们在Google上搜索信息时, Google会推荐一些相关联的搜索词, 即使我们拼写错了关键词, Google仍然会推荐出我们想要的内容, 这就是搜索推荐.

Elasticsearch Suggesters

Elasticsearch Suggesters 包含四个功能, 下面分别介绍这四个功能及其Java Client的使用.

Term Suggester

Term Suggester 即推荐相似的term. 原理是对输入的搜索文本进行分词得到term, 然后再去倒排索引中找到相似的term返回, 如果存在相同的term, 就返回为空.

语法

POST 索引名/_search
{
  "suggest": {
    "任意名称": {
      "text": "搜索文本",
      "term": {
        "field": "字段",
        "analyzer": "",
        "size": "",
        "sort": "",
        "suggest_mode": "",
        "prefix_length":"",
        ...
      }
    }
  }
}

(1) analyzer

搜索文本使用的分词器.

(2) size

最多返回几个推荐term.

(3) sort

推荐term的排序.

• score
  相关度分数优先.
• frequency
  相似term的个数优先.

(4) suggest_mode

term的推荐规则.

• missing
  如果倒排索引中存在相同的term就不推荐, 默认配置.
• popular
  只推荐比原来term出现频率更高的term.
• always
  不管倒排索引中是否存在相同的term都要推荐.(当然如果没有相似的term就返回空)

(5) prefix_length

匹配前缀长度, 默认为1. 默认情况下必须匹配首字母才会进行推荐.

(6) 其他

此外还有很多参数, 可以去官网上查看.

https://www.elastic.co/guide/en/elasticsearch/reference/current/search-suggesters.html#term-suggester

实例

(1) 添加数据

POST articles/_bulk
{ "index" : { } }
{ "title": "lucene", "content": "lucene is very cool"}
{ "index" : { } }
{ "title": "elasticsearch", "content": "Elasticsearch builds on top of lucene"}
{ "index" : { } }
{ "title": "Elasticsearch", "content": "Elasticsearch rock"}
{ "index" : { } }
{ "title": "ELK", "content": "elastic is the company behind ELK stack"}
{ "index" : { } }
{ "title": "Elk", "content": "Elk stack rocks"}
{ "index" : {} }
{ "title": "es", "content": "elasticsearch is rock solid"}

(2) 实例一

POST articles/_search
{
  "suggest": {
    "my_suggestion_01": {
      "text": "lucen rocks",
      "term": {
        "field": "content"
      }
    },
    "my_suggestion_02": {
      "text": "wucene",
      "term": {
        "prefix_length": 0,
        "field": "title"
      }
    }
  }
}

结果:

"suggest" : {
	"my_suggestion_01" : [
	  {
	    "text" : "lucen",
	    "offset" : 0,
	    "length" : 5,
	    "options" : [
	      {
	        "text" : "lucene",
	        "score" : 0.8,
	        "freq" : 2
	      }
	    ]
	  },
	  {
	    "text" : "rocks",
	    "offset" : 6,
	    "length" : 5,
	    "options" : [ ]
	  }
	],
	"my_suggestion_02" : [
	  {
	    "text" : "wucene",
	    "offset" : 0,
	    "length" : 6,
	    "options" : [
	      {
	        "text" : "lucene",
	        "score" : 0.8333333,
	        "freq" : 1
	      }
	    ]
	  }
	]
}

(3) 实例二

POST articles/_search
{
  "suggest": {
    "text": "lucen rocks",
    "my_suggestion_01": {
      "term": {
	    "suggest_mode": "always",
           "field": "content"
      }
    },
    "my_suggestion_02": {
      "term": {
        "field": "title"
      }
    }
  }
}

结果:

"suggest" : {
	"my_suggestion_01" : [
	  {
	    "text" : "lucen",
	    "offset" : 0,
	    "length" : 5,
	    "options" : [
	      {
	        "text" : "lucene",
	        "score" : 0.8,
	        "freq" : 2
	      }
	    ]
	  },
	  {
	    "text" : "rocks",
	    "offset" : 6,
	    "length" : 5,
	    "options" : [
	      {
	        "text" : "rock",
	        "score" : 0.75,
	        "freq" : 2
	      }
	    ]
	  }
	],
	"my_suggestion_02" : [
	  {
	    "text" : "lucen",
	    "offset" : 0,
	    "length" : 5,
	    "options" : [
	      {
	        "text" : "lucene",
	        "score" : 0.8,
	        "freq" : 1
	      }
	    ]
	  },
	  {
	    "text" : "rocks",
	    "offset" : 6,
	    "length" : 5,
	    "options" : [ ]
	  }
	]
}

Phrase Suggester

Phrase Suggester 即推荐相似的短语. 它在Term Suggester的基础上, 考量多个term之间的关系(比如是否同时出现在索引的原文里, 相邻程度以及词频等), 推荐一个相似的短语.

语法

POST 索引名/_search
{
  "suggest": {
    "任意名称": {
      "text": "搜索文本",
      "phrase": {
        "field": "字段",
        "analyzer": "",
        "size": "",
        "max_errors": "",
        "highlight":"",
        ...
      }
    }
  }
}

(1) analyzer

搜索文本使用的分词器.

(2) size

最多返回几个推荐短语, 默认5个, 常用的值为3或5.

(3) max_errors

最多可以拼写错的term数, 默认1, 常用值为1或2.

(4) highlight

将短语中推荐的term高亮显示.

(5) 其他

此外还有很多参数, 可以去官网上查看.

https://www.elastic.co/guide/en/elasticsearch/reference/current/search-suggesters.html#phrase-suggester

实例

(1) 添加数据

POST articles/_bulk
{ "index" : { } }
{ "title"