How to highlight ngram tokens in a word using elastic search - elasticsearch

Question

I would like to highlight just the ngrams which match, not the whole word.
Example:
term: "Wo"
highlight should be: "<em>Wo</em>nderfull world!"
currently it is: "<em>Wonderfull</em> world!"
Mapping is:
{
"global_search_1495732922733" : {
"mappings" : {
"meeting" : {
"properties" : {
...
"name" : {
"type" : "text",
"analyzer" : "meeteor_index_analyzer",
"search_analyzer" : "meeteor_search_term_analyzer"
},
...
}
}
}
}
}
Analyzers are:
"analysis" : {
"filter" : {
"meeteor_stemmer" : {
"name" : "english",
"type" : "stemmer"
},
"meeteor_ngram" : {
"type" : "nGram",
"min_gram" : "2",
"max_gram" : "15"
}
},
"analyzer" : {
"meeteor_search_term_analyzer" : {
"filter" : [
"lowercase",
"asciifolding"
],
"tokenizer" : "standard"
},
"meeteor_index_analyzer" : {
"filter" : [
"lowercase",
"asciifolding",
"meeteor_ngram"
],
"tokenizer" : "standard"
},
"meeteor_project_id_analyzer" : {
"tokenizer" : "standard"
}
}
},
Concrete example:
curl -XGET 'localhost:9200/global_search/meeting/_search?pretty' -H 'Content-Type: application/json' -d'
{
"query": {
"match": {
"name": "Me"
}
},
"highlight":{
"fields": {
"name": {}
}
}
}
'
The result is:
"...highlight" : {
"name" : [
"Sad <em>Meeting</em>"
]
}

Answer 1

The correct way to achieve what you want is using ngram as tokenizer and not filter. You can do something like this:
"analysis" : {
"filter" : {
"meeteor_stemmer" : {
"name" : "english",
"type" : "stemmer"
}
},
"tokenizer" : {
"meeteor_ngram_tokenizer" : {
"type" : "nGram",
"min_gram" : "2",
"max_gram" : "15"
}
},
"analyzer" : {
"meeteor_search_term_analyzer" : {
"filter" : [
"lowercase",
"asciifolding"
],
"tokenizer" : "standard"
},
"meeteor_index_analyzer" : {
"filter" : [
"lowercase",
"asciifolding"
],
"tokenizer" : "meeteor_ngram_tokenizer"
},
"meeteor_project_id_analyzer" : {
"tokenizer" : "standard"
}
}
},
It will generate the highlighting by ngram for you like this:
"...highlight" : {
"name" : [
"Sad <em>Me</em>eting"
]
}