Elasticsearch 5.5：名称搜索需要帮助（职位得分较高）答案

【问题标题】：Elasticsearch 5.5: Needs help for name search (Higher score for positions)Elasticsearch 5.5：名称搜索需要帮助（职位得分较高）
【发布时间】：2017-07-12 19:24:09
【问题描述】：

我正在尝试进行自动补全，它会为键入的单词位置提供更高的分数。按分数排序，然后按名称排序。

我的目标：

输入“宠物”

结果：

peter christensen
peter christian grau
peter christian reumert krogsgaard
peter bruun christensen
anders petersen

输入“peter chr”

结果：

peter christensen
peter christian grau
peter christian reumert krogsgaard
peter bruun christensen

上面成功了，我很高兴，但是当有重复的“peter christensen”时发生了一些奇怪的事情。现在结果如下所示：

peter christian grau
peter christian reumert krogsgaard
peter christensen
peter christensen
peter bruun christensen

我想要这个：

peter christensen
peter christensen
peter christian grau
peter christian reumert krogsgaard
peter bruun christensen

请帮忙？

设置：

{
    "persons_index": {
        "settings": {
            "index": {
                "number_of_shards": "5",
                "provided_name": "persons_index",
                "creation_date": "1499881803116",
                "analysis": {
                    "filter": {
                        "ascii_folding_preserve_original": {
                            "type": "asciifolding",
                            "preserve_original": "true"
                        },
                        "names_synonym_filter": {
                            "type": "synonym",
                            "synonyms": [
                                "aage,åge",
                                "gaard,gård"
                            ]
                        }
                    },
                    "analyzer": {
                        "my_analyzer": {
                            "filter": [
                                "lowercase",
                                "trim",
                                "names_synonym_filter",
                                "ascii_folding_preserve_original"
                            ],
                            "type": "custom",
                            "tokenizer": "standard"
                        }
                    }
                },
                "number_of_replicas": "1",
                "uuid": "LxMHLha-R02i__S7gDWBtA",
                "version": {
                    "created": "5050099"
                }
            }
        }
    }
}

映射：

{
    "persons_index": {
        "mappings": {
            "persons_type": {
                "properties": {
                    "fieldDisplayFullName": {
                        "type": "text",
                        "norms": false,
                        "analyzer": "my_analyzer"
                    },
                    "fieldSort": {
                        "type": "text",
                        "norms": false,
                        "analyzer": "keyword",
                        "fielddata": true
                    }
                }
            }
        }
    }
}

示例文档

{"fieldSort" : "peter christensen", "fieldDisplayFullName" : "Peter Christensen"}

查询：

_search?pretty=true&search_type=dfs_query_then_fetch

{
  "explain": false,
  "size" : 50,
  "sort": [
        "_score",
        {
            "fieldSort": {
                "order": "asc"
            }
        }
    ],
  "query": {
    "bool" : {
      "must" : {
        "bool" : {
              "minimum_should_match":"2",
              "should" : [
                {"match" : { "fieldDisplayFullName" : "peter" }},
                {"wildcard" : { "fieldDisplayFullName" : "chr*" }}
              ]
        }
      },
       "should" : [
          {"match_phrase_prefix" : { "fieldSort" : {"query" : "peter chr","boost" : 10}}}
        ]
    }
  }
}

结果：

Peter Christensen (_score: 18.968887)
Peter Christian Engelhardt (_score: 18.968887)
Peter Christian Grau (_score: 18.968887)
Peter Christian Reumert Krogsgaard (_score: 18.968887)
Peter Christian Vagnbo Jørgensen (_score: 18.968887)
Peter Christoffersen (_score: 18.968887)
Peter Bruun Christensen (_score: 1.0512933)
Peter Dits Christensen (_score: 1.0512933)
Peter Fjeldsø Christensen (_score: 1.0512933)

重复“彼得克里斯滕森（_score：14.909464）”

Peter Christian Engelhardt (_score: 20.01772)
Peter Christian Grau (_score: 20.01772)
Peter Christian Reumert Krogsgaard (_score: 20.01772)
Peter Christian Vagnbo Jørgensen (_score: 20.01772)
Peter Christoffersen (_score: 20.01772)
Peter Christensen (_score: 14.909464)
Peter Christensen (_score: 14.909464)
Peter Bruun Christensen (_score: 1.04652)
Peter Dits Christensen (_score: 1.04652)
Peter Fjeldsø Christensen (_score: 1.04652)

不同的查询尝试没有成功：

{
  "explain": false,
  "size" : 50,
  "sort": [
        "_score",
        {
            "fieldSort": {
                "order": "asc"
            }
        }
    ],
  "query": {
    "bool" : {
      "must" : {
        "bool" : {
              "minimum_should_match":"2",
              "should" : [
                {"match" : { "fieldDisplayFullName" : "peter" }},
                {"wildcard" : { "fieldDisplayFullName" : "chr*" }}
              ]
        }
      },
       "should" : [
            {"span_first" : {
              "match" : {
                  "span_term" : { "fieldDisplayFullName" : "peter" }
              },
              "end" : 1,
              "boost" : 50.0
            }},
            {"span_first" : {
              "match" : {
                  "span_multi":{
                    "match":{
                        "wildcard" : { "fieldDisplayFullName" : "chr*" }
                    }
                  }
              },
              "end" : 2,
              "boost" : 50.0
            }}
        ]
    }
  }
}

【问题讨论】：

标签： elasticsearch

【解决方案1】：

用“constant_score”解决

{
  "explain": false,
  "size" : 50,
  "sort": [
        "_score",
        {
            "fieldSort": {
                "order": "asc"
            }
        }
    ],
  "query": {
    "bool" : {
      "must" : {
        "bool" : {
              "minimum_should_match":"2",
              "should" : [
                {"match" : { "fieldDisplayFullName" : "peter" }},
                {"wildcard" : { "fieldDisplayFullName" : "chr*" }}
              ]
        }
      },
       "should" : [
            { "constant_score": {
                "boost":   2,
                "query": {"match_phrase_prefix" : { "fieldSort" : {"query" : "peter chr"}}}
            }}
        ]
    }
  }
}

【讨论】：

感谢您发布解决方案！保重，多米尼克 :)

【解决方案2】：

看看 Elasticsearch 如何computes relevance scores 尤其是IDF（逆文档频率）。

因此，您可以尝试多个approaches，通过忽略 IDF 来解决此问题。我有一个类似的问题，我通过将应用的search type 更改为dfs_query_then_fetch-queries 解决了这个问题。

您的配置是否在超过 1 个shard 上运行？

干杯，多米尼克

【讨论】：

嗨多米尼克。感谢您的回答。我刚刚用更多信息更新了我的问题。我已经在使用 search_type=dfs_query_then_fetch 了。分片数是 5？我将深入研究您今天晚些时候提供给我的链接并返回。谢谢杰斯珀
感谢 Dominik 为我指明了正确的方向。阅读您的“方法”链接后，我花了 5 分钟。用“constant_score”解决。