CODEWITHSUNDEEP
phpelasticsearch
Lawrence Cooke
Lawrence Cooke

Reputation: 1597

Elasticsearch - full text search

I am trying to match "new york" in a search (not places containing "new" or "york" separately)

here is my current query:

"query" : {                        
    "query_string" : {            
        "query" : "new york" ,
        "fields" : ["city"]
    }
},
"filter" : {
    "and" : [{
        "query" : {                 
            "query_string" : {                    
                "query" : "country:US"               
             }            
        }
    }]
}

However this keeps returning places named "york" rather than "new york"

I am not fully understanding how this works and would appreciate some help in getting this to actually work for me.

Upvotes: 0

Views: 2489

Answers (2)

Moghira
Moghira

Reputation: 317

By default the city field in "analyzed" by Elasticsearch which does the default tokenization of words.

New York => *New*, *York*

In order to keep the tokens intact (mostly used for aggregation), you need to explicitly make the city field "not analyzed" using multi field

'city' => [
            'type' => 'string',
            'fields' => [
            'raw' => [
              'type'  => 'string',
              'index' => 'not_analyzed'
              ]
            ]
        ]

Now you can use city.raw to get non analyzed values.

Upvotes: 0

javanna
javanna

Reputation: 60245

If you want both words to appear in the same document you need to change the default operator like this:

"query" : {                        
    "query_string" : {            
        "query" : "new york" ,
        "fields" : ["city"],
        "default_operator" : "AND"
    }
}

or specify it ion the query:

"query" : {                        
    "query_string" : {            
        "query" : "new AND york" ,
        "fields" : ["city"]
    }
}

Have a look at the query string documentation.

Otherwise, if you want both words to appear close to each other in the same document you need to make a phrase query like this:

"query" : {
    "match_phrase" : {
        "message" : "new york"
    }
}

Upvotes: 3

Related Questions