elasticsearch - 为什么我得到不同字段的综合得分而不是最佳匹配字段的得分

Question

我试图通过将multi_match与best_fields以及一个自定义分析器(对每个字段应用带状阴影和/或ngram)一起使用，从不同的mutifield文档中找到最匹配的字段。我希望得到最佳匹配字段的分数，但是我会得到最佳匹配文档不同字段的综合分数。为什么会这样呢？有其他方法吗？有人可以告诉我如何获得最佳匹配 Realm 的分数吗？
这是我正在做的一个例子:

PUT /my-index1
{
  "mappings": {
    "_doc": {
      "properties": {
        "name1": {
          "type": "text",
          "fields": {
            "name_shingles": {
              "type": "text",
              "analyzer": "shingle_analyz",
              "search_analyzer": "shingle_analyz"
            }
          }
        },
        "name2": {
          "type": "text",
          "fields": {
            "name_shingles": {
              "type": "text",
              "analyzer": "shingle_analyz",
              "search_analyzer": "shingle_analyz"
            }
          }
        }
      }
    }
  },
  "settings": {
    "index": {
      "number_of_shards": "1",
      "analysis": {
        "filter": {
          "shingle_filter": {
            "max_shingle_size": "3",
            "min_shingle_size": "2",
            "output_unigrams": false,
            "type": "shingle"
          },
          "duplicate_token_filter": {
            "type": "unique",
            "only_on_same_position": false
          }
        },
        "analyzer": {
          "shingle_analyz": {
            "type": "custom",
            "tokenizer": "standard",
            "filter": [
              "lowercase",
              "asciifolding",
              "shingle_filter",
              "duplicate_token_filter"
            ],
            "char_filter": [
              "my_char_filter"
            ]
          }
        },
        "char_filter": {
          "my_char_filter": {
            "type": "mapping",
            "mappings": [
              "'s => s",
              "'S => S"
            ]
          }
        }
      },
      "number_of_replicas": "1"
    }
  }
}
 

PUT /my-index1/_doc/1
{
  "name1": "WA WA WA WOMAN",
  "name2": "wa wa"
}

PUT /my-index1/_doc/2
{
  "name1": "WA WA WOMAN",
  "name2": "help"
}

PUT /my-index1/_doc/3
{
  "name1": "the great showman",
  "name2": "WA WA WOMAN"
}


GET my-index1/_search
{
  "explain": true,
  "query": {
    "bool": {
      "should": [
        {
          "multi_match": {
            "fields": [
              "name1.name_shingles",
              "name2.name_shingles"
            ],
            "query": "WA WA WOMAN",
            "type": "best_fields",
            "operator": "or"
          }
        }
      ]
    }
  }
}

Answer 1

经过一些研究，我得出的结论是，在处理带状板和多个字段时，match和muti_match查询会产生不可靠的分数。我尝试了许多不同的配置，并且总是混淆了不同 Realm 的分数。