Search match fulltext find with uncomplete word but not found with complete word in Elasticsearch

Question

I have a really weird problem in my elasticsearch query. I made an autocomplete search in my website and I have a problem.

For example, there is a neighborhood in my country called "Recreio dos Bandeirantes" When I search for "bandeirant" (while user are typing) the query find the neighborhood, but, when finish the type "bandeirantes" cannot find the same neighborhood.

This is my query

      {
        query: {
          bool: {
            must: [
              {
                match: {
                  'city.name': city,
                },
              },
              {
                match: {
                  'city.state': state,
                },
              },
              {
                match: {
                  keyword: {
                    query, // The query is 'bandeirant' or 'bandeirantes'
                  },
                },
              },
            ],
          },
        },
        highlight: {
          fields: {
            keyword: {
              number_of_fragments: 9,
            },
          },
        },
        size: 20,
      }

The final neighborhood value is 'Recreio dos Bandeirantes, Rio de Janeiro, RJ'

The mapping for this field is this:

{
  "search-neighborhood-01": {
    "mappings": {
      "properties": {
        "city": {
          //.....
        },
        "keyword": {
          "type": "text",
          "analyzer": "autocomplete",
          "search_analyzer": "standard"
        },
        "name": {
          "type": "text",
          "analyzer": "autocomplete",
          "search_analyzer": "standard"
        }
      }
    }
  }
}

My settings with analyzer

{
  "search-neighborhood-01": {
    "settings": {
      "index": {
        // .......
        "analysis": {
          "filter": {
            "autocomplete_filter": {
              "token_chars": [
                "letter"
              ],
              "min_gram": "1",
              "type": "edge_ngram",
              "max_gram": "10"
            }
          },
          "analyzer": {
            "autocomplete": {
              "filter": [
                "lowercase",
                "autocomplete_filter",
                "asciifolding"
              ],
              "type": "custom",
              "tokenizer": "standard"
            }
          }
        },
        // .....
      }
    }
  }
}

My response with bandeirant

// .....
      {
       //.....
        "_source": {
          "city": {
            "name": "Rio de Janeiro",
            "state": "RJ",
            "keyword": "Rio de Janeiro, RJ"
          },
          "name": "Recreio dos Bandeirantes",
          "keyword": "Recreio dos Bandeirantes, Rio de Janeiro, RJ"
        },
        "highlight": {
          "keyword": [
            "Recreio dos <em>Bandeirantes</em>, Rio de Janeiro, RJ"
          ]
        }
      }

My response with bandeirantes is empty :/

How can I do to solve this?

Thanks o/

Answer 1

I changed my max_ngram to 20 and worked :)

Answer 2

You have this issue because you Ngram filter token have"max_gram": "10" configuration that it means words longer than 10 won’t be indexed.

My recommendation would be to increase this amount along with “min_gram” config.