I have a pretty complex query and now I want to boost some documents that fulfill some criteria. I have the following simplified document structure and I try to give some documents a boost based on the id, genre, tag.
{
"id": 123,
"genres": ["ACTION", "DRAMA"],
"tags": ["For kids", "Romantic", "Nature"]
}
What I want to do is for example
id: 123 boost: 5
genres: ACTION boost: 3
tags: Romantic boost: 0.2
and boost all documents that are contained in my query and fit the criteria but I don't want to filter them out. So query clause boosting is not of any help I guess.
Edit: To make if easier to understand what I want to achieve (not sure if it is possible with elasticsearch, no is also a valid answer).
I want to search with a query and get a result set. In this set I want to boost some documents. But I don't want to enlarge the result set or filter it. The boost should be independent from the query.
For example I search for a specific tag and want to boost all documents with category 'ACTION' in the result set. But I don't want all documents with category 'ACTION' in the result set and also I don't want only documents with the specific tag AND category 'ACTION'.
I think you need to have Dynamic boosting during query time.
The first matches the id title with boost and second one matches the 'genders' ACTION.
{
"query": {
"bool": {
"should": [
{
"match": {
"title": {
"query": "id",
"boost": 5
}
}
},
{
"match": {
"content": "Action"
}
}
]
}
}
}
If you want to have multi_match match based on your query:
{
"multi_match" : {
"query": "some query terms here",
"fields": [ "id^5", "genders^3", "tags^0.2" ]
}
}
Note: the ^5 means boost for the title.
Edit:
Maybe you are asking for different types of multi_match queries (at least for ES 5.x) from the ES reference guide:
best_fields
(default) Finds documents which match any field, but uses
the _score from the best field. See best_fields.
most_fields
Finds documents which match any field and combines the _score from
each field. See most_fields.
cross_fields
Treats fields with the same analyzer as though they were one big
field. Looks for each word in any field. See cross_fields.
phrase
Runs a match_phrase query on each field and combines the _score from
each field. See phrase and phrase_prefix.
phrase_prefix
Runs a match_phrase_prefix query on each field and combines the _score
from each field. See phrase and phrase_prefix.
More at: ES 5.4 ElasticSearch reference
I found a solution and it was pretty simple. I use a boosting query. I now just nest the different boosting criteria with and my original query is now the base query.
https://www.elastic.co/guide/en/elasticsearch/reference/2.3/query-dsl-boosting-query.html
For example:
{
"query": {
"boosting": {
"positive": {
"boosting": {
"positive": {
"match": {
"director": "Spielberg"
}
},
"negative": {
"term": {
"genres": "DRAMA"
}
},
"negative_boost": 1.3
}
},
"negative": {
"term": {
"tags": "Romantic"
}
},
"negative_boost": 1.2
}
}
}
Related
I'm a little confused about what is the difference between should and boost final score calculation
when a bool query has a must clause, the should clauses act as a boost factor, meaning none of them have to match but if they do, the relevancy score for that document will be boosted and thus appear higher in the result.
so,if we have:
one query which contains must and should clauses
vs
second query which contains must clause and boosting clause
Is there a difference ?
when you recommend to use must and should vs must and boosting clauses in a query ?
You can read the documentation of boolean query here, there is huge difference in the should and boost.
Should and must both contributes to the _score of the document, and as mentioned in the above documentation, follows the
The bool query takes a more-matches-is-better approach, so the score from each matching must or should clause will be added together to provide the final _score for each document.
While boost is a parameter, using which you can increase the weight according to your value, let me explain that using an example.
Index sample docs
POST _doc/1
{
"brand" : "samsung",
"name" : "samsung phone"
}
POST _doc/2
{
"brand" : "apple",
"name" : "apple phone"
}
Boolean Query using should without boost
{
"query": {
"bool": {
"should": [
{
"match": {
"name": {
"query": "apple"
}
}
},
{
"match": {
"brand": {
"query": "apple"
}
}
}
]
}
}
}
Search result showing score
"max_score": 1.3862942,
Now in same query use boost of factor 10
{
"query": {
"bool": {
"should": [
{
"match": {
"name": {
"query": "apple"
}
}
},
{
"match": {
"brand": {
"query": "apple",
"boost": 10 --> Note additional boost
}
}
}
]
}
}
}
Query result showing boost
"max_score": 7.624619, (Note considerable high score)
In short, when you want to boost a particular document containing your query term, you can additionally pass the boost param and it will be on top of the normal score calculated by should or must.
I have an Elastic search index that contain thousands of documents, each document represent a user.
each document has set of fields (is_verified: boolean, country: string, is_creator: boolean), also i have another service that call ES search to lookup for documents, how i can rank the retrieved documents based on those fields? for example a verified user with match should come first than un verified one.
is there some kind of document scoring while indexing the documents ? if yes can i modify it based on my criteria ?
what shall i read/look to understand how to rank in elastic search.
thanks
I guess the sorting function mentioned by Mikael is pretty straight forward and should cover your use cases. Check Elastic Doc for more information on that.
But in case you want to do really fancy sorting, maybe you could use a bool query and different boost values to set your desired relevancy for each matched field. It tried to come up with a real life example, but honestly didn't find one. For the sake of completeness, he following snippet should give you an idea how to achieve similar results as with the sort API (but still, i would prefer using sort).
GET /yourindexname/_search
{
"query": {
"bool": {
"must": [
{
"match": {
"name": "Monica"
}
}
],
"should": [
{
"term": {
"is_verified": {
"value": true,
"boost": 2
}
}
},
{
"term": {
"is_creator": {
"value": true,
"boost": 2
}
}
}
]
}
}
}
is there some kind of document scoring while indexing the documents ? if yes can i modify it based on my criteria ?
I wouldn't assign a fixed score to a document while indexing, as the score should be dependent on the query. However, if you insist to have a predefined relevancy for each document, theoretically you could add a field relevancy having that value for ordering and use it later in the query:
GET /yourindexname/_search
{
"query" : {
"match" : {
"name": "Monica"
}
},
"sort" : [
{
"relevancy": {
"order": "desc"
},
"_score"
}
]
}
You can consider using the Sort Api inside your search queries ,In example below we used the search on the field country and sorted the result with respect of Boolean field (is_verified) , You can also add the other Boolean field inside Sort brackets .
GET /yourindexname/_search
{
"query" : {
"match" : {
"country": "Iceland"
}
},
"sort" : [
{
"is_verified": {
"order": "desc"
}
}
]
}
I'm a beginner in ElasticSearch, I have an application that uses elasticSearch to look for ingredients in a given food or fruit...
I'm facing a problem with scoring if the user for example tapes: "Vitamine d"
ElasticSearch will give the "vitamine" phrase that has the best scoring even if the phrase "Vitamine D" exists and normally it should have the highest score.
I see that if the second word "d" in my case is just one letter then elastic search will ignore it.
I did another example: "vitamine b12" and I had the correct score.
Here is the query that the application send to the server:
{
"from": 0,
"size": 5,
"query": {
"bool": {
"must": [
{
"match": {
"constNomFr": {
"query": "vitamine d"
}
}
}
],
"should": [
{
"prefix": {
"constNomFr": {
"value": "vitamine d",
"boost": 2
}
}
}
]
}
},
"_source": {
"excludes": [
"alimentDtos"
]
}
}
What could I modify to make it work?
Thank you so much.
If you can identify your ingredients, I recommend you to index them on a separate field "ingredients" setting it's type to keyword. This way you can use a term filter and you can even run aggregations.
You may already have your documents indexed that way, in that case if your are using the default mapping, just run your query against your_field_name.keyword.
If you don't have your ingredients indexed as an array then you should take a look to the elasticsearch analyzers to choose or build the right one.
I have a database with entries like
title: This is my awesome title
abstract: A more detailed descriptions of what [...]
I would like to build an Elasticsearch query that matches the above document with, e.g.,
awe detai
In words: A multi_match phrase_prefix query with multiple search terms. (This is intended to be used as a search-as-you-type feature.)
I see how you can combine multi_match and phrase_prefix, but it's unclear to me how to do this for multiple search terms.
Any hints?
Well there is few ways to do that
POST stack/autocomplete/1
{
"title": "This is my awesome title",
"abstract": "A more detailed descriptions of what"
}
Then you can search using query string with star but problem here is that you need to append asterix to query
POST stack/autocomplete/_search
{
"query": {
"query_string": {
"fields": [
"title",
"abstract"
],
"query": "awe* detai*"
}
}
}
If you want to match on user query then you can use like that
POST stack/autocomplete/_search
{
"query": {
"multi_match": {
"fields": [
"title",
"abstract"
],
"query": "awesome tit",
"type": "phrase_prefix"
}
}
}
One more option to consider would be to use nGram with query string so you will not need to modify user query "awe* detai*"
What is the difference between such query:
"query": {
"bool": {
...
"should": [
{
"match": {
"description": {
"query": "test"
}
}
},
{
"match": {
"address": {
"query": "test",
}
}
},
{
"match": {
"country": {
"query": "test"
}
}
},
{
"match": {
"city": {
"query": "test"
}
}
}
]
}}
and that one:
"query": {
"bool": {
...
"should": [
{
"query_string": {
"query": "test",
"fields": [
"description",
"address",
"country",
"city"
]
}
}
]
}}
Performance, relevance?
Thanks in advance!
The query is analyzed depending on the field analyzer (unless you specify the analyzer in the query itself), thus querying multiple fields with a single query doesn't necessarily mean analyzing the query only once.
Keep in mind that the query_string supports the lucene query syntax: AND and OR operators, querying on specific fields, wildcard, phrase queries etc. therefore it needs to be parsed, which I don't think makes a lot of difference here in terms of performance, but it is error prone and might lead to errors. If you don't need all that power, stick to the match query, and if you want to perform the same query on multiple fields, have a look at the multi_match query, which does what you did with your query_string but translates internally to multiple match queries.
Also, the scores returned if you compare the output of multiple match queries and your query_string might be quite different. Using a bool query you effectively build a lucene boolean query, while the query_string uses by default "use_dis_max":"true", which means it uses internally a dis_max query by default. Same happens using the multi_match query. If you set use_dis_max to false a bool query is going to be used internally instead.
I terms of performance, I would say that the second query will have performance benefits because, the first query requires the query string to be analyzed for all the four match sections, while in the second there is only one query string that needs to be analyzed.
Apart from that, there are some comparisons done over here that you can look at.
I am not quite sure about the relevancy differences, but that you can always fire these two queries and see if there is any difference in relevance from the results fetched.