Elasticsearch：嵌套文档匹配评分无法累加问题求助

我有一组代表候选人的文档，每个文档包含嵌套的workExperiences工作经历子文档。需求是每个匹配的工作经历都为候选人的评分贡献权重，因此使用了权重为6的function_score函数，但多个工作经历匹配时评分并未如预期累加——例如两个匹配项时预期得分为12（2*6），但实际仍为6分。

现有查询语句 #

GET /candidates/_search
{
  "query": {
    "nested": {
      "path": "workExperiences",
      "query": {
        "function_score": {
          "query": {
            "match": {
              "workExperiences.name.raw": "software engineer"
            }
          },
          "functions": [
            {
              "filter": {
                "match": {
                  "workExperiences.name.raw": "software engineer"
                }
              },
              "weight": 6
            }
          ],
          "score_mode": "sum"
        }
      }
    }
  },
  "aggs": {
    "total": {
      "cardinality": {
        "field": "id"
      }
    }
  },
  "sort": [
    {
      "_score": {
        "order": "desc"
      }
    },
    {
      "id": {
        "order": "desc"
      }
    }
  ],
  "track_total_hits": true,
  "explain": true,
  "from": 0,
  "size": 10,
  "collapse": {
    "field": "id"
  },
  "_source": [
    "id"
  ]
}

映射配置 #

{
  "candidates_0": {
    "mappings": {
      "properties": {
        "workExperiences": {
          "type": "nested",
          "include_in_root": true,
          "properties": {
            "createdAt": {
              "type": "date"
            },
            "name": {
              "type": "text",
              "fields": {
                "raw": {
                  "type": "text",
                  "analyzer": "raw_analyzer"
                }
              }
            }
          }
        }
      }
    }
  }
}

索引设置 #

{
  "candidates_0": {
    "settings": {
      "index": {
        "routing": {
          "allocation": {
            "include": {
              "_tier_preference": "data_content"
            }
          }
        },
        "number_of_shards": "5",
        "provided_name": "candidates_0",
        "creation_date": "170809684784",
        "analysis": {
          "filter": {
            "trim_filter": {
              "type": "trim"
            }
          },
          "analyzer": {
            "raw_analyzer": {
              "filter": [
                "lowercase",
                "asciifolding",
                "trim_filter"
              ],
              "type": "custom",
              "tokenizer": "keyword"
            }
          }
        },
        "number_of_replicas": "1",
        "uuid": "aVhyAoJfTgyqUk0QbMJFKA",
        "version": {
          "created": "8444511"
        }
      }
    }
  }
}

问题原因 #

核心问题在于默认的nested查询只会取所有匹配嵌套文档中的最高得分作为父文档的最终得分，不会累加所有匹配项的分数。当前的function_score内部设置了score_mode: sum，但这个sum仅作用于当前单个嵌套文档内的函数得分计算，无法影响父文档对多个匹配嵌套文档的得分汇总逻辑。

另外，functions中的filter属于冗余配置——外层的match已经过滤出符合条件的嵌套文档，无需重复过滤。

解决方案 #

在nested查询中添加score_mode: sum参数，让父文档的得分等于所有匹配嵌套文档的得分之和，同时简化冗余配置：

修正后的查询语句 #

GET /candidates/_search
{
  "query": {
    "nested": {
      "path": "workExperiences",
      "score_mode": "sum", // 关键：累加所有匹配嵌套文档的得分
      "query": {
        "function_score": {
          "query": {
            "match": {
              "workExperiences.name.raw": "software engineer"
            }
          },
          "functions": [
            {
              "weight": 6 // 每个匹配的嵌套文档贡献6分
            }
          ],
          "score_mode": "sum",
          "boost_mode": "replace" // 可选：直接使用权重值作为嵌套文档得分，替代原始匹配得分
        }
      }
    }
  },
  "aggs": {
    "total": {
      "cardinality": {
        "field": "id"
      }
    }
  },
  "sort": [
    {
      "_score": {
        "order": "desc"
      }
    },
    {
      "id": {
        "order": "desc"
      }
    }
  ],
  "track_total_hits": true,
  "explain": true,
  "from": 0,
  "size": 10,
  "collapse": {
    "field": "id"
  },
  "_source": [
    "id"
  ]
}

关键修正点说明 #

1. nested查询的score_mode: sum：这是实现多嵌套文档得分累加的核心配置，让父文档的得分是所有匹配嵌套文档得分的总和。
2. 移除冗余filter：外层match已过滤出目标嵌套文档，每个匹配项都会应用weight:6的权重。
3. 可选boost_mode: replace：如果希望嵌套文档的得分直接等于权重值（而非权重与原始匹配得分相乘），可以添加此配置，确保每个匹配项稳定贡献6分。

内容的提问来源于stack exchange，提问作者Denis Walter