如何在Elasticsearch中实现emailId字段的子文本高亮？

在Elasticsearch中，companyName和country字段可正常高亮匹配用户搜索关键词的子文本，但emailId字段却会高亮整个字段而非匹配的子串。以下是相关信息及解决方法：

查询语句 #

POST customers-index/_search
{
  "size": 20,
  "query": {
    "must": [
      {
        "multi_match": {
          "query": "state",
          "fields": [
            "emailId.autocomplete",
            "companyName.autocomplete",
            "country.autocomplete"
          ]
        }
      }
    ]
  },
  "highlight": {
    "fields": {
      "emailId.autocomplete": {},
      "companyName.autocomplete": {},
      "country.autocomplete": {}
    }
  },
  "_source": [
    "emailId",
    "companyName",
    "country"
  ]
}

响应结果 #

[
 {
        "_index" : "customers-index",
        "_type" : "_doc",
        "_id" : "c44aefde-22b1-471d-9159-a092e5c604f6",
        "_score" : 14.853605,
        "_source" : {
          "country" : "Ethiopia",
          "companyName" : "La Galtoara, Inc.",
          "emailId" : [
            "galtoara@state.gov"
          ]
        },
        "highlight" : {
          "emailId.autocomplete" : [
            "<em>galtoara@state.gov</em>"
          ]
        }
      },
      {
        "_index" : "customers-index",
        "_type" : "_doc",
        "_id" : "f76ecf0a-3e7d-41f9-a96f-83c66698f2d1",
        "_score" : 3.6045084,
        "_source" : {
          "country" : "Philippines",
          "companyName" : "Belgone State Medical, Inc.",
          "emailId" : [
            "dopasdfd@apple.com"
          ]
        },
        "highlight" : {
          "companyName.autocomplete" : [
            "Belgone <em>State</em> Medical, Inc."
          ]
        }
      },
      {
        "_index" : "customers-index",
        "_type" : "_doc",
        "_id" : "b41b1c0c-e84d-4424-a862-38b10d380d23",
        "_score" : 2.1431046,
        "_source" : {
          "country" : "United States",
          "companyName" : "DFDFDF Brands Limited",
          "emailId" : [
            "adfadfad@godaddy.com"
          ]
        },
        "highlight" : {
          "country.autocomplete" : [
            "United <em>State</em>s"
          ]
        }
      }
]

emailId字段映射 #

{
  "emailId": {
    "type": "text",
    "fields": {
      "autocomplete": {
        "type": "text",
        "analyzer": "autocomplete_email_analyzer",
        "search_analyzer": "search_analyzer"
      },
      "keyword": {
        "type": "keyword",
        "ignore_above": 256,
        "normalizer": "lowercase_normalizer"
      }
    },
    "analyzer": "index_analyzer",
    "search_analyzer": "search_analyzer"
  }
}

索引配置 #

{
  "analysis": {
    "filter": {
      "email_filter": {
        "type": "pattern_capture",
        "preserve_original": "true",
        "patterns": [
          "\"(?=([@|\\.|\\!|\\#|\\$|%|&|'|\\*|\\+|\\-|\\/|\\=|\\?|\\^|\\_|\\`|\\{|\\||\\}|\\~](.+)))\""
        ]
      },
      "starts_with_filter": {
        "type": "edge_ngram",
        "min_gram": "1",
        "max_gram": "100"
      }
    },
    "analyzer": {
      "search_analyzer": {
        "filter": [
          "lowercase"
        ],
        "tokenizer": "keyword"
      },
      "index_analyzer": {
        "filter": [
          "lowercase"
        ],
        "tokenizer": "index_analyzer"
      },
      "autocomplete_email_analyzer": {
        "filter": [
          "email_filter",
          "unique",
          "starts_with_filter",
          "lowercase"
        ],
        "tokenizer": "autocomplete_email_tokenizer"
      }
    },
    "normalizer": {
      "lowercase_normalizer": {
        "filter": [
          "lowercase"
        ],
        "type": "custom",
        "char_filter": []
      }
    },
    "tokenizer": {
      "autocomplete_email_tokenizer": {
        "type": "uax_url_email"
      },
      "index_analyzer": {
        "token_chars": [
          "letter",
          "digit",
          "whitespace",
          "punctuation",
          "symbol"
        ],
        "min_gram": "2",
        "type": "ngram",
        "max_gram": "30"
      }
    }
  }
}

问题原因 #

1. 搜索分析器不匹配：emailId.autocomplete使用的search_analyzer采用keyword分词器，会将搜索关键词"state"作为完整token处理，高亮时会匹配整个email字段而非子串。
2. 邮箱拆分过滤器异常：email_filter的正则表达式存在错误，无法正确拆分邮箱中的@、.分隔的子部分，导致索引时的token无法精准对应搜索关键词。

解决步骤 #

1. 修正邮箱拆分过滤器 #

更新索引配置中的email_filter，使用正确的正则表达式捕获邮箱的各个子部分：

"email_filter": {
  "type": "pattern_capture",
  "preserve_original": "true",
  "patterns": [
    "(@)(.+)",
    "(\\.)(.+)"
  ]
}

该配置会保留原始邮箱字符串，同时捕获@之后的域名部分和.之后的后缀部分，配合edge ngram生成更精准的索引token。

2. 替换搜索分析器 #

创建适合邮箱搜索的分析器，替代原有的search_analyzer：

"analyzer": {
  // 保留其他原有分析器
  "search_email_analyzer": {
    "filter": ["lowercase"],
    "tokenizer": "uax_url_email"
  }
}

然后修改emailId.autocomplete的映射，使用新的搜索分析器：

"autocomplete": {
  "type": "text",
  "analyzer": "autocomplete_email_analyzer",
  "search_analyzer": "search_email_analyzer"
}

uax_url_email分词器会正确拆分邮箱为多个部分，让搜索时的token能匹配索引中的子串token。

3. 优化高亮配置 #

在查询的highlight部分指定unified类型，确保高亮器能精准定位匹配子串：

"highlight": {
  "fields": {
    "emailId.autocomplete": {
      "type": "unified",
      "fragment_size": 100
    },
    "companyName.autocomplete": {},
    "country.autocomplete": {}
  }
}

4. 重新索引数据 #

修改索引配置和映射后，需要重新索引所有数据，确保新的分析器和过滤器生效。

完成以上步骤后，emailId字段就能像companyName和country一样，仅高亮匹配搜索关键词的子文本。

内容的提问来源于stack exchange，提问作者Karthik Amar