ElasticSearch 中现有字段的完成建议

Question

在我的 elasticsearch 索引中，我索引了一堆工作。为简单起见，我们只说它们是一堆职称。当人们在我的搜索引擎中输入职位名称时，我想 "Auto Complete" 找到可能的匹配项。

我在这里调查了 Completion Suggester：http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/search-suggesters-completion.html

然而，我发现的所有示例都涉及在您的索引上创建一个新字段，并在 indexing/rivering 时手动填充该字段。

有什么方法可以在现有字段上获得完成建议吗？即使这意味着重新索引数据也没关系。例如，当我想保留原始 not_analysed 文本时，我可以在映射中做这样的事情：

"JobTitle": {
    "type": "string",
    "fields": {
        "Original": {
            "type": "string",
            "index": "not_analyzed"
        }
    }
}

这可能与建议者有关吗？

如果不是，是否可以进行非空白 tokenizing/N-Gram 搜索来获取这些字段？虽然它会更慢，但我认为这会起作用。

Answer 1

好的，这是（可能或）可能无法缩放的简单方法，使用 prefix queries。

我将使用您提到的 "fields" 技术和我找到的一些方便的职位描述数据创建索引 here:

DELETE /test_index

PUT /test_index
{
   "settings": {
      "number_of_shards": 1,
      "number_of_replicas": 0
   },
   "mappings": {
      "doc": {
         "properties": {
            "title": {
               "type": "string",
               "fields": {
                  "raw": {
                     "type": "string",
                     "index": "not_analyzed"
                  }
               }
            }
         }
      }
   }
}

PUT /test_index/_bulk
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"experienced bra fitter", "desc":"I bet they had trouble finding candidates for this one."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"PlayStation Brand Ambassador", "desc":"please report to your residence in the United States of Nintendo."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Eyebrow Threading", "desc":"I REALLY hope this has something to do with dolls."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Administraive/ Secretary", "desc":"ok, ok, we get it. It’s clear where you need help."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Finish Carpenter", "desc":"for when the Start Carpenter gets tired."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Helpdesk Technician @ Pentagon", "desc":"“Uh, hello? I’m having a problem with this missile…”"}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Nail Tech", "desc":"so nails can be pretty complicated…"}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Remedy Engineer", "desc":"aren’t those called “doctors”?"}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Saltlick Cashier", "desc":"new trend in the equestrian industry. Ok, enough horsing around."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Molecular Biologist II", "desc":"when Molecular Biologist I gets promoted."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Breakfast Sandwich Maker", "desc":"we also got one of these recently."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Hotel Housekeepers", "desc":"why can’t they just say ‘hotelkeepers’?"}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Preschool Teacher #4065", "desc":"either that’s a really big school or they’ve got robot teachers."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"glacéau drop team", "desc":"for a new sport at the Winter Olympics: ice-water spilling."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"PLUMMER/ELECTRICIAN", "desc":"get a dictionary/thesaurus first."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"DoodyCalls Technician", "desc":"they really shouldn’t put down janitors like that."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Golf Staff", "desc":"and here I thought they were called clubs."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Pressure Washers", "desc":"what’s next, heat cleaners?"}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Sandwich Artist", "desc":"another “Jesus in my food” wannabe."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Self Storage Manager", "desc":"this is for self storage?"}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Qualified Infant Caregiver", "desc":"too bad for all the unqualified caregivers on the list."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Ground Support", "desc":"but there’s just more dirt under there."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Gymboree Teacher", "desc":"the hardest part is not burning your hands sliding down the pole."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"COMMERCIAL space hunter", "desc":"so they did find animals further out in the cosmos? Who knew."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"JOB COACH", "desc":"if they’re unemployed when they get to you, what does that say about them?"}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"KIDS KAMP INSTRUCTOR!", "desc":"no spelling ability required."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"POOLS SUPERVISOR", "desc":"“yeah, they’re still wet…”"}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"HOUSE MANAGER/TEEN SUPERVISOR", "desc":"see the dictionary under P, for Parent."}
{"index":{"_index":"test_index","_type":"doc"}}
{"title":"Licensed Seamless Gutter Contractor", "desc":"just sounds bad."}

那我就可以轻松运行一个前缀查询：

POST /test_index/_search
{
    "query": {
        "prefix": {
           "title": {
              "value": "san"
           }
        }
    }
}
...
{
   "took": 6,
   "timed_out": false,
   "_shards": {
      "total": 1,
      "successful": 1,
      "failed": 0
   },
   "hits": {
      "total": 2,
      "max_score": 1,
      "hits": [
         {
            "_index": "test_index",
            "_type": "doc",
            "_id": "mcRfqtwzTyWE7ZNsKFvwEg",
            "_score": 1,
            "_source": {
               "title": "Breakfast Sandwich Maker",
               "desc": "we also got one of these recently."
            }
         },
         {
            "_index": "test_index",
            "_type": "doc",
            "_id": "fIYV0WOWRe6gfpYy_u2jlg",
            "_score": 1,
            "_source": {
               "title": "Sandwich Artist",
               "desc": "another “Jesus in my food” wannabe."
            }
         }
      ]
   }
}

或者如果我想对匹配更加小心，我可以使用未分析的字段：

POST /test_index/_search
{
    "query": {
        "prefix": {
           "title.raw": {
              "value": "San"
           }
        }
    }
}
...
{
   "took": 1,
   "timed_out": false,
   "_shards": {
      "total": 1,
      "successful": 1,
      "failed": 0
   },
   "hits": {
      "total": 1,
      "max_score": 1,
      "hits": [
         {
            "_index": "test_index",
            "_type": "doc",
            "_id": "fIYV0WOWRe6gfpYy_u2jlg",
            "_score": 1,
            "_source": {
               "title": "Sandwich Artist",
               "desc": "another “Jesus in my food” wannabe."
            }
         }
      ]
   }
}

这是简单的方法。 Ngrams 有点复杂，但并不困难。我稍后会在另一个答案中添加它。

这是我使用的代码：

http://sense.qbox.io/gist/4e066d051d7dab5fe819264b0f4b26d958d115a9

编辑：Ngram 版本

借用this blog post的解析器（不要脸的插件），我可以这样设置索引：

DELETE /test_index

PUT /test_index
{
   "settings": {
      "number_of_shards": 1,
      "number_of_replicas": 0,
      "analysis": {
         "filter": {
            "nGram_filter": {
               "type": "nGram",
               "min_gram": 2,
               "max_gram": 20,
               "token_chars": [
                  "letter",
                  "digit",
                  "punctuation",
                  "symbol"
               ]
            }
         },
         "analyzer": {
            "nGram_analyzer": {
               "type": "custom",
               "tokenizer": "whitespace",
               "filter": [
                  "lowercase",
                  "asciifolding",
                  "nGram_filter"
               ]
            },
            "whitespace_analyzer": {
               "type": "custom",
               "tokenizer": "whitespace",
               "filter": [
                  "lowercase",
                  "asciifolding"
               ]
            }
         }
      }
   },
   "mappings": {
      "doc": {
         "properties": {
            "title": {
               "type": "string",
               "index_analyzer": "nGram_analyzer", 
               "search_analyzer": "whitespace_analyzer", 
               "fields": {
                  "raw": {
                     "type": "string",
                     "index": "not_analyzed"
                  }
               }
            }
         }
      }
   }
}

请注意，我使用不同的分析器进行索引和搜索；这很重要，因为如果搜索查询被分解成 ngram，我们可能会得到比我们想要的更多的点击率。

使用上面使用的相同数据集进行填充，我可以使用简单的 match 查询来获得我期望的结果：

POST /test_index/_search
{
    "query": {
        "match": {
           "title": "sup"
        }
    }
}
...
{
   "took": 1,
   "timed_out": false,
   "_shards": {
      "total": 1,
      "successful": 1,
      "failed": 0
   },
   "hits": {
      "total": 3,
      "max_score": 1.8631258,
      "hits": [
         {
            "_index": "test_index",
            "_type": "doc",
            "_id": "4pcAOmPNSYupjz7lSes8jw",
            "_score": 1.8631258,
            "_source": {
               "title": "Ground Support",
               "desc": "but there’s just more dirt under there."
            }
         },
         {
            "_index": "test_index",
            "_type": "doc",
            "_id": "DVFOC6DsTa6eH_a-RtbUUw",
            "_score": 1.8631258,
            "_source": {
               "title": "POOLS SUPERVISOR",
               "desc": "“yeah, they’re still wet…”"
            }
         },
         {
            "_index": "test_index",
            "_type": "doc",
            "_id": "klleY_bnQ4uFmCPF94sLOw",
            "_score": 1.4905007,
            "_source": {
               "title": "HOUSE MANAGER/TEEN SUPERVISOR",
               "desc": "see the dictionary under P, for Parent."
            }
         }
      ]
   }
}

代码如下：

http://sense.qbox.io/gist/b0e77bb7f05a4527de5ab4345749c793f923794c

ElasticSearch 中现有字段的完成建议

Completion Suggester in ElasticSearch On Existing Field

lucene

elasticsearch