Elasticsearch generate suggestion fields - elasticsearch

Question

I've been reading in to the suggestion in elasticsearch in blogs like: https://www.elastic.co/blog/you-complete-me
But there you have to put in the name_suggest data your self, isn't there a way to automaticly add the data to the name_suggest when you map the object.
so update this mapping:
curl -X PUT localhost:9200/hotels -d '
{
"mappings": {
"hotel" : {
"properties" : {
"name" : { "type" : "string" },
"city" : { "type" : "string" },
"name_suggest" : {
"type" : "completion"
}
}
}
}
}'
and with these puts:
curl -X PUT localhost:9200/hotels/hotel/1 -d '
{
"name" : "Mercure Hotel Munich",
"city" : "Munich",
"name_suggest" : "Mercure Hotel Munich"
}'
curl -X PUT localhost:9200/hotels/hotel/2 -d '
{
"name" : "Hotel Monaco",
"city" : "Munich",
"name_suggest" : "Hotel Monaco"
}'
curl -X PUT localhost:9200/hotels/hotel/3 -d '
{
"name" : "Courtyard by Marriot Munich City",
"city" : "Munich",
"name_suggest" : "Courtyard by Marriot Munich City"
}'
so we can lose the name_suggest field.
So the ultimate goal is when you start typing Ho the first result would be Hotel

Answer 1

You can do it with ngrams if you want partial matches within words, or edge ngrams if you just want to match from the beginning of words.
Here's an example. I set up an index like this:
PUT /test_index
{
"settings": {
"analysis": {
"filter": {
"edge_ngram_filter": {
"type": "edge_ngram",
"min_gram": 2,
"max_gram": 20
}
},
"analyzer": {
"edge_ngram_analyzer": {
"type": "custom",
"tokenizer": "standard",
"filter": [
"lowercase",
"edge_ngram_filter"
]
}
}
}
},
"mappings": {
"doc": {
"properties": {
"name": {
"type": "string",
"index_analyzer": "edge_ngram_analyzer",
"search_analyzer": "standard"
},
"city": {
"type": "string"
}
}
}
}
}
Then added your docs:
POST /test_index/doc/_bulk
{"index":{"_id":1}}
{"name":"Mercure Hotel Munich","city":"Munich"}
{"index":{"_id":2}}
{"name":"Hotel Monaco","city":"Munich"}
{"index":{"_id":3}}
{"name":"Courtyard by Marriot Munich City","city":"Munich"}
Now I can query for documents with "hot" in the name like this:
POST /test_index/_search
{
"query": {
"match": {
"name": "hot"
}
}
}
and I get back the correct docs:
{
"took": 41,
"timed_out": false,
"_shards": {
"total": 1,
"successful": 1,
"failed": 0
},
"hits": {
"total": 2,
"max_score": 0.625,
"hits": [
{
"_index": "test_index",
"_type": "doc",
"_id": "2",
"_score": 0.625,
"_source": {
"name": "Hotel Monaco",
"city": "Munich"
}
},
{
"_index": "test_index",
"_type": "doc",
"_id": "1",
"_score": 0.5,
"_source": {
"name": "Mercure Hotel Munich",
"city": "Munich"
}
}
]
}
}
There are various ways this can be tweaked or generalized. For example, you can apply the ngram analyzer to the _all field if you want to match on more than one field.
Here is the code I used to test it:
http://sense.qbox.io/gist/3583de02c4f7d33e07ba4c2def9badf90692a290