Giving priority to prefix match in elasticsearch in php - php

Question

Is there a way in elasticsearch to give more priority for the prefix match than to the string that contains that word?
For ex.- priorities of words if I search for ram should be like this:
Ram Reddy
Joy Ram Das
Kiran Ram Goel
Swati Ram Goel
Ramesh Singh
I have tried mapping as given in here.
I have done like this:
$params = [
"index" => $myIndex,
"body" => [
"settings"=> [
"analysis"=> [
"analyzer"=> [
"start_with_analyzer"=> [
"tokenizer"=> "my_edge_ngram",
"filter"=> [
"lowercase"
]
]
],
"tokenizer"=> [
"my_edge_ngram"=> [
"type"=> "edge_ngram",
"min_gram"=> 3,
"max_gram"=> 15
]
]
]
],
"mappings"=> [
"doc"=> [
"properties"=> [
"label"=> [
"type"=> "text",
"fields"=> [
"keyword"=> [
"type"=> "keyword"
],
"ngramed"=> [
"type"=> "text",
"analyzer"=> "start_with_analyzer"
]
]
]
]
]
]
]
];
$response = $client->indices()->create($params); // create an index
and searching like this:
$body = [
"size" => 100,
'_source' => $select,
"query"=> [
"bool"=> [
"should"=> [
[
"query_string"=> [
"query"=> "ram*",
"fields"=> [
"value"
],
"boost"=> 5
]
],
[
"query_string"=> [
"query"=> "ram*",
"fields"=> [
"value.ngramed"
],
"analyzer"=> "start_with_analyzer",
"boost"=> 2
]
]
],
"minimum_should_match"=> 1
]
]
];
$params = [
'index' => $myIndex,
'type' => $myType,
'body' => []
];
$params['body'] = $body;
$response = $client->search($params);
The json of query is as follows:
{
"size": 100,
"_source": [
"label",
"value",
"type",
"sr"
],
"query": {
"bool": {
"should": [
{
"query_string": {
"query": "ram*",
"fields": [
"value"
],
"boost": 5
}
},
{
"query_string": {
"query": "ram*",
"fields": [
"value.ngramed"
],
"analyzer": "start_with_analyzer",
"boost": 2
}
}
],
"minimum_should_match": 1,
"must_not": {
"match_phrase": {
"type": "propertyValue"
}
}
}
}
}
I am using elasticsearch 5.3.2
Is there any other way to sort the results for the search in the relational database using the search method in php?

Answer 1

You should not enable fielddata unless really required. To overcome this you can use sub field.
Make the following changes to your code:
"label"=>[
"type"=>"text",
//"fielddata"=> true, ---->remove/comment this line
"analyzer"=>"whitespace",
"fields"=>[
"keyword"=>[
"type"=>"keyword"
]
]
]
To sort on type field use type.keyword instead. This change apply to any field of text type and has a sub-field of type keyword available (assuming the name of this field is keyword). So change as below:
'sort' => [
["type.keyword"=>["order"=>"asc"]],
["sr"=>["order"=>"asc"]],
["propLabels"=>["order"=>"asc"]],
["value"=>["order"=>"asc"]]
]
Update : Index creation and query to get desired output
Create the index as below:
{
"settings": {
"analysis": {
"analyzer": {
"start_with_analyzer": {
"tokenizer": "my_edge_ngram",
"filter": [
"lowercase"
]
}
},
"tokenizer": {
"my_edge_ngram": {
"type": "edge_ngram",
"min_gram": 3,
"max_gram": 15
}
}
}
},
"mappings": {
"_doc": {
"properties": {
"name": {
"type": "text",
"fields": {
"keyword": {
"type": "keyword"
},
"ngramed": {
"type": "text"
}
}
}
}
}
}
}
Use the query below to get the desired result:
{
"query": {
"bool": {
"should": [
{
"query_string": {
"query": "Ram",
"fields": [
"name"
],
"boost": 5
}
},
{
"query_string": {
"query": "Ram",
"fields": [
"name.ngramed"
],
"analyzer": "start_with_analyzer",
"boost": 2
}
}
],
"minimum_should_match": 1
}
}
}
In the above the query with boost value 5 increases the score for those documents where Ram is present in name. The other query with boost 2 further increases the score for the documents where name starts with Ram.
Sample O/P:
"hits": [
{
"_index": "test",
"_type": "_doc",
"_id": "2",
"_score": 2.0137746,
"_source": {
"name": "Ram Reddy"
}
},
{
"_index": "test",
"_type": "_doc",
"_id": "1",
"_score": 1.4384104,
"_source": {
"name": "Joy Ram Das"
}
},
{
"_index": "test",
"_type": "_doc",
"_id": "3",
"_score": 0.5753642,
"_source": {
"name": "Ramesh Singh"
}
}
]