Elastic Search: “Exact” phrase matching with wildcards

问题

I'm using Elastic to search for names in a genealogical database. One of the options for the search is "exact search". The problem is that my clients want wildcards to be allowed in the exact search, so the difference between exact and inexact is that inexact search will return fuzzy matches, whereas exact should return the exact phrase searched for with the exception of wildcards (no fuzzy results).

In order to enable wildcards, the search is currently using querystring. This is the format of the exact search:

{
  "query": {
      "filtered": {
          "query": {
              "bool": {
                  "must": [
                      {
                          "dis_max": {
                              "queries": [
                                  {
                                      "match": {
                                          "first_name": {
                                              "type": "phrase",
                                              "query": "mary c.",
                                              "fuzziness": 0,
                                              "analyzer": "standard",
                                              "boost": 2
                                          }
                                      }
                                  },
                                  {
                                      "query_string": {
                                          "query": "mary c.",
                                          "default_field": "first_name",
                                          "analyzer": "standard",
                                          "fuzzy_min_sim": 0,
                                          "boost": 0.5
                                      }
                                  }
                              ]
                          }
                      }
                  ]
              }
          }
      }
  }

}

I have a boost so that fully exact matches are returned first, which works fine. However, after my exact matches, I get (using Mary C. as an example) results like "Mary F." or "James C.". My clients don't want this, as it's not exact enough; I should ONLY get results with name Mary C., or, if I search for "Mar* C." I should get "Mary C." or "Martin C.", but I shouldn't get "James C." or "Mary F."

I added "default_operator": "AND" to the querystring like such:

{
    "query_string": {
        "query": "mary c.",
        "default_field": "first_name",
        "analyzer": "standard",
        "fuzzy_min_sim": 0,
        "boost": 0.5,
        "default_operator": "AND"
    }
}

which is better, but still not quite right; now I only get results that have "Mary" AND "C." in the first name, but some of them are "Mary Jane C." and "Mary, widow of James C."

Is there any way I can make the query_string match more exactly? At the very least, the phrase prefix should match, so "Mary C." shouldn't return "Mary, widow of James C." but only "Mary C. ....". Ideally, Mary C. should ONLY match "Mary C." and "Mar* C." will match "Mary C.", "Martin C.", etc.

来源：https://stackoverflow.com/questions/27275213/elastic-search-exact-phrase-matching-with-wildcards

标签

ElasticSearch

match

wildcard