CODEWITHSUNDEEP
elasticsearch
pierallard
pierallard

Reputation: 3371

Elasticsearch - Index word, bigram and trigram

I'm trying to index some phrases, like this :

"Elasticsearch is a great search engine"

indexed like

Elasticsearch       # word
is                  # word
a                   # word
great               # word
engine              # word
Elasticsearch is    # bi-gram
is a                # bi-gram
a great             # bi-gram
great search        # bi-gram
search engine       # bi-gram
Elasticsearch is a  # tri-gram
is a great          # tri-gram
a great search      # tri-gram
great search engine # tri-gram

I know how to index words (with default indexer) and to index bigrams and trigrams (with n-grams indexer), but not both at same time.

How can I do this ?

Regards

Upvotes: 3

Views: 3457

Answers (1)

Nathan Smith
Nathan Smith

Reputation: 8347

You would use the multi-field type. Here is an example of one I have created - 

{
  "settings": {
    "number_of_shards": 5,
    "number_of_replicas": 0,
    "analysis": {
      "filter": {
        "synonym": {
          "type": "synonym",
          "synonyms_path": "synonyms.txt"
        },
        "my_metaphone": {
          "type": "phonetic",
          "encoder": "metaphone",
          "replace": false
        }
      },
      "analyzer": {
        "synonym": {
          "tokenizer": "whitespace",
          "filter": [
            "lowercase",
            "synonym"
          ]
        },
        "metaphone": {
          "tokenizer": "standard",
          "filter": [
            "my_metaphone"
          ]
        },
        "porter": {
          "tokenizer": "standard",
          "filter": [
            "lowercase",
            "porter_stem"
          ]
        }
      }
    }
  },
  "mappings": {
    "type": {
      "_all": {
        "enabled": false
      },
      "properties": {
        "datafield": {
          "type": "multi_field",
          "store": "yes",
          "fields": {
            "datafield": {
              "type": "string",
              "analyzer": "simple"
            },
            "metaphone": {
              "type": "string",
              "analyzer": "metaphone"
            },
            "porter": {
              "type": "string",
              "analyzer": "porter"
            },
            "synonym": {
              "type": "string",
              "analyzer": "synonym"
            }
          }
        }
      }
    }
  }
}

You can then specify which field you want to search against, i.e. datafield.synonym or in your case datafield.bigram. You can then build your query up, boosting the fields as to which is most important to your results.

Upvotes: 3

Related Questions