CODEWITHSUNDEEP
elasticsearch
Ronnie
Ronnie

Reputation: 551

Contains Query or Partial matching in Elasticsearch

I'm looking to search a word say "amend" which may be present in data as "amending", "amendment" or even "*amend". Which is the best method to search words like these? I know wildcard can achieve this but I am restricted to not using it due to my other part of the code. What are the different ways which provides better search performance?

Upvotes: 15

Views: 48026

Answers (3)

Tim
Tim

Reputation: 1286

What you're describing with the terms "amend", "amendment" and "amending" is called keyword stemming. You can add a stemmer token filter to your Elastic index settings.

For Example:

PUT /my_index
{
    "settings": {
        "analysis" : {
            "analyzer" : {
                "my_analyzer" : {
                    "tokenizer" : "standard",
                    "filter" : ["standard", "lowercase", "my_stemmer"]
                }
            },
            "filter" : {
                "my_stemmer" : {
                    "type" : "stemmer",
                    "name" : "english"
                }
            }
        }
    }
}

Using this stemmer will index the terms [amend, amending, amendment] as [amend, amend, amend].

Then you can do a match query and it should return what you're wanting.

Upvotes: 1

Ashish Tiwari
Ashish Tiwari

Reputation: 2277

There is various way:

As you mention you cannot use wildcard, Then go for query_string

{  
   "query":{  
      "query_string":{  
         "default_field":"text",
         "query":"*amend"
      }
   }
}

Second you can use n-gram tokenizer. You can check here https://www.elastic.co/guide/en/elasticsearch/reference/current/analysis-ngram-tokenizer.html

It will breaks your value "amending" to words like ["ame","men","end" ... etc]

Once you done with applying n-gram tokeinzer, Start indexing your data.

You can query like below:

{"query":{"term":{"text":"amend"}}}

You will get your output result.

Upvotes: 5

Kamal Kunjapur
Kamal Kunjapur

Reputation: 8840

You can implement this using query_string feature of elasticsearch. Assuming that you use default standard analyzer.

{  
   "query":{  
      "query_string":{  
         "default_field":"Customer",
         "query":"*Jo*"
      }
   }
}

You can add multiple fields as well as shown in the below query

{  
   "query":{  
      "query_string":{  
         "fields":[  
            "Customer",
            "Name"
         ],
         "query":"*Jo*"
      }
   }
}

Upvotes: 18

Related Questions