Some Multi word synonyms are not working in elasticsearch for nested fields - elasticsearch

Question

I am trying to use synonym analyzer at query time and not getting expected results. Can someone throw some light on this?
Here is my mapping for the index:
{
"jobs_user_profile_v2": {
"mappings": {
"profile": {
"_all": {
"enabled": false
},
"_ttl": {
"enabled": true
},
"properties": {
"rsa": {
"type": "nested",
"properties": {
"answer": {
"type": "string",
"index_analyzer": "autocomplete",
"search_analyzer": "synonym",
"position_offset_gap": 100
},
"answerId": {
"type": "long"
},
"answerOriginal": {
"type": "string",
"index": "not_analyzed"
},
"createdAt": {
"type": "long"
},
"label": {
"type": "string",
"index": "not_analyzed"
},
"labelOriginal": {
"type": "string",
"index": "not_analyzed"
},
"question": {
"type": "string",
"index": "not_analyzed"
},
"questionId": {
"type": "long"
},
"questionOriginal": {
"type": "string"
},
"source": {
"type": "integer"
},
"updatedAt": {
"type": "long"
}
}
}
}
}
}
}
}
The field to focus on is rsa.answer, which is the field I am querying.
My synonym mapping:
Beautician,Stylist,Make up artist,Massage therapist,Therapist,Spa,Hair Dresser,Salon,Beauty Parlour,Parlor => Beautician
Carpenter,Wood Worker,Furniture Carpenter => Carpenter
Cashier,Store Manager,Store Incharge,Purchase Executive,Billing Executive,Billing Boy => Cashier
Content Writer,Writer,Translator,Writing,Copywriter,Content Creation,Script Writer,Freelance Writer,Freelance Content Writer => Content Writer
My Search Query:
http://{{domain}}/jobs_user_profile_v2/_search
{
"query": {
"nested":{
"path": "rsa",
"query":{
"query_string": {
"query": "hair dresser",
"fields": ["answer"],
"analyzer" :"synonym"
}
},
"inner_hits": {
"explain": true
}
}
},
"explain" : true,
"sort" : [ {
"_score" : { }
} ]
}
It is showing proper Beautician and 'Cashierprofiles for search queryHair Dresserandbilling executivebut not showing anything forwood worker => carpenter` case.
My analyzer results:
http://{{domain}}/jobs_user_profile_v2/_analyze?analyzer=synonym&text=hair dresser
{
"tokens": [
{
"token": "beautician",
"start_offset": 0,
"end_offset": 12,
"type": "SYNONYM",
"position": 1
}
]
}
and for wood worker case
http://{{domain}}/jobs_user_profile_v2/_analyze?analyzer=synonym&text=wood worker
{
"tokens": [
{
"token": "carpenter",
"start_offset": 0,
"end_offset": 11,
"type": "SYNONYM",
"position": 1
}
]
}
It is also not working a few other cases.
My analyzer setting for index:
"analysis": {
"filter": {
"synonym": {
"ignore_case": "true",
"type": "synonym",
"synonyms_path": "synonym.txt"
},
"autocomplete_filter": {
"type": "edge_ngram",
"min_gram": "3",
"max_gram": "10"
}
},
"analyzer": {
"text_en_splitting_search": {
"type": "custom",
"filter": [
"stop",
"lowercase",
"porter_stem",
"word_delimiter"
],
"tokenizer": "whitespace"
},
"synonym": {
"filter": [
"stop",
"lowercase",
"synonym"
],
"type": "custom",
"tokenizer": "standard"
},
"autocomplete": {
"filter": [
"lowercase",
"autocomplete_filter"
],
"type": "custom",
"tokenizer": "standard"
},
"text_en_splitting": {
"filter": [
"lowercase",
"porter_stem",
"word_delimiter"
],
"type": "custom",
"tokenizer": "whitespace"
},
"text_general": {
"filter": [
"lowercase"
],
"type": "custom",
"tokenizer": "standard"
},
"edge_ngram_analyzer": {
"filter": [
"lowercase"
],
"type": "custom",
"tokenizer": "edge_ngram_tokenizer"
},
"autocomplete_analyzer": {
"filter": [
"lowercase"
],
"tokenizer": "whitespace"
}
},
"tokenizer": {
"edge_ngram_tokenizer": {
"token_chars": [
"letter",
"digit"
],
"min_gram": "2",
"type": "edgeNGram",
"max_gram": "10"
}
}
}

Answer 1

For the above case one multi-match is more ideal than query-string.
Multi-Match unlike query string does not tokenize the query terms before analyzing it . As a result multi-word synonyms may not work as expected.
Example:
{
"query": {
"nested": {
"path": "rsa",
"query": {
"multi_match": {
"query": "wood worker",
"fields": [
"rsa.answer"
],
"type" : "cross_fields",
"analyzer": "synonym"
}
}
}
}
}
If for some reason you prefer query-string then you would need to pass the entire query in double quotes to ensure it is not tokenized:
example :
post test/_search
{
"query": {
"nested": {
"path": "rsa",
"query": {
"query_string": {
"query": "\"wood worker\"",
"fields": [
"rsa.answer"
],
"analyzer": "synonym"
}
}
}
}
}