如何配置ElasticSearch使其计算词频权重时兼顾双字段？

针对你遇到的ArticleTest类型文档因标签匹配权重机制，始终排在匹配度更高的ArticleBasic文档之前的问题，有以下几种可配置的解决方案：

1. 给不同类型的标签查询设置差异化权重（Boost） #

直接在ArticleBasic的标签匹配查询中添加boost参数，提升这类查询的得分权重，让匹配更多标签的ArticleBasic文档能获得更高的总分，超过ArticleTest文档。

修改后的查询示例：

{
  "query": {
    "bool": {
      "should": [
        { "term": { "contenttype": "ArticleTest" } },
        { "term": { "contenttype": "ArticleBasic" } }
      ],
      "must": [
        {
          "bool": {
            "should": [
              { "wildcard": { "ArticleBasic.tags": "*history*", "boost": 2 } },
              { "wildcard": { "ArticleBasic.tags": "*meet locals*", "boost": 2 } },
              { "wildcard": { "ArticleBasic.tags": "*nature & wildlife*", "boost": 2 } },
              { "wildcard": { "ArticleBasic.tags": "*|north-america|canada|quebec-city*", "boost": 2 } },
              { "wildcard": { "ArticleBasic.tags": "*|north-america|canada|churchill*", "boost": 2 } },
              { "wildcard": { "ArticleBasic.tags": "*|north-america|canada|manitoba*", "boost": 2 } },
              { "wildcard": { "ArticleBasic.tags": "*|north-america|canada|alberta*", "boost": 2 } },
              { "wildcard": { "ArticleTest.tags": "*history*" } },
              { "wildcard": { "ArticleTest.tags": "*meet locals*" } },
              { "wildcard": { "ArticleTest.tags": "*nature & wildlife*" } },
              { "wildcard": { "ArticleTest.tags": "*|north-america|canada|quebec-city*" } },
              { "wildcard": { "ArticleTest.tags": "*|north-america|canada|churchill*" } },
              { "wildcard": { "ArticleTest.tags": "*|north-america|canada|manitoba*" } },
              { "wildcard": { "ArticleTest.tags": "*|north-america|canada|alberta*" } }
            ]
          }
        }
      ],
      "filter": [
        { "bool": { "must_not": [{ "match": { "identifier": "b05e28402b854b9221f08daa81e99b61" } }] } }
      ]
    }
  },
  "sort": [
    { "_score": "desc" },
    {
      "_script": {
        "type": "number",
        "script": "if(doc['ArticleTest.publish'].size() != 0) { return doc['ArticleTest.publish'].value.millis; }else if(doc['ArticleBasic.publish'].size() != 0) { return doc['ArticleBasic.publish'].value.millis; }else { return 0; }",
        "order": "desc"
      }
    }
  ],
  "size": 5,
  "from": 0
}

这里给所有ArticleBasic的标签查询设置了boost:2，意味着每个匹配的得分会是原来的2倍，你可以根据实际情况调整这个数值。

2. 使用function_score自定义得分逻辑 #

如果需要更灵活的权重计算，可以用function_score查询，结合文档类型和标签匹配情况自定义得分。比如给ArticleBasic文档设置基础权重乘数，同时保留标签匹配的得分累加：

示例查询：

{
  "query": {
    "function_score": {
      "query": {
        "bool": {
          "must": [
            {
              "bool": {
                "should": [
                  { "wildcard": { "ArticleBasic.tags": "*history*" } },
                  { "wildcard": { "ArticleBasic.tags": "*meet locals*" } },
                  { "wildcard": { "ArticleBasic.tags": "*nature & wildlife*" } },
                  { "wildcard": { "ArticleBasic.tags": "*|north-america|canada|quebec-city*" } },
                  { "wildcard": { "ArticleBasic.tags": "*|north-america|canada|churchill*" } },
                  { "wildcard": { "ArticleBasic.tags": "*|north-america|canada|manitoba*" } },
                  { "wildcard": { "ArticleBasic.tags": "*|north-america|canada|alberta*" } },
                  { "wildcard": { "ArticleTest.tags": "*history*" } },
                  { "wildcard": { "ArticleTest.tags": "*meet locals*" } },
                  { "wildcard": { "ArticleTest.tags": "*nature & wildlife*" } },
                  { "wildcard": { "ArticleTest.tags": "*|north-america|canada|quebec-city*" } },
                  { "wildcard": { "ArticleTest.tags": "*|north-america|canada|churchill*" } },
                  { "wildcard": { "ArticleTest.tags": "*|north-america|canada|manitoba*" } },
                  { "wildcard": { "ArticleTest.tags": "*|north-america|canada|alberta*" } }
                ]
              }
            }
          ],
          "filter": [
            { "bool": { "must_not": [{ "match": { "identifier": "b05e28402b854b9221f08daa81e99b61" } }] } },
            { "bool": { "should": [
              { "term": { "contenttype": "ArticleTest" } },
              { "term": { "contenttype": "ArticleBasic" } }
            ] } }
          ]
        }
      },
      "functions": [
        {
          "filter": { "term": { "contenttype": "ArticleBasic" } },
          "weight": 1.5
        },
        {
          "filter": { "term": { "contenttype": "ArticleTest" } },
          "weight": 1.0
        }
      ],
      "score_mode": "multiply"
    }
  },
  "sort": [
    { "_score": "desc" },
    {
      "_script": {
        "type": "number",
        "script": "if(doc['ArticleTest.publish'].size() != 0) { return doc['ArticleTest.publish'].value.millis; }else if(doc['ArticleBasic.publish'].size() != 0) { return doc['ArticleBasic.publish'].value.millis; }else { return 0; }",
        "order": "desc"
      }
    }
  ],
  "size": 5,
  "from": 0
}

这里通过weight参数给ArticleBasic文档设置了1.5倍的得分乘数，ArticleTest保持1倍，最终得分是标签匹配的原始得分乘以这个权重。score_mode设置为multiply表示权重与原始得分相乘，你也可以根据需求选择sum等其他模式。

3. 优化标签字段的索引（长期方案） #

当前使用wildcard查询标签字段效率低，且得分计算不够精准。建议将tags字段拆分为多值keyword字段：

• 修改索引映射，将ArticleBasic.tags和ArticleTest.tags设置为type: keyword且multi: true（或者预先按|拆分存储为数组）。
• 查询时使用terms或match查询替代wildcard，这样每个标签匹配的得分计算更准确，也更容易通过boost调整权重。

例如，拆分后的标签字段查询可以改为：

{ "terms": { "ArticleBasic.tags": ["history", "meet locals", "nature & wildlife", "north-america|canada|quebec-city"], "boost": 2 } }

这种方式不仅查询效率更高，得分计算也更符合预期。

内容的提问来源于stack exchange，提问作者Jmh2013