Elasticsearch mapping config file boost field - search

Question

I'm new to search and am having trouble interpreting the documentation on boosting fields in the mapping.
I want to achieve a simple boosting where the title of some article is more important than the tags associated with the article.
Here's an attempt at the config, which I have put in config/[index_name]/[some_name].json:
{
"[type]": {
"properties": {
"_boost": {
"name": "title",
"null_value": 2.0
}
"title": {
"type": "string"
}
}
}
}
I can tell the file is being read because of error messages from previous attempts at this file. I have also been deleting the index and recreating it between attempts so that it will use this mapping.
Will this work? It doesn't give any error messages, but I can't tell if there is any boost in effect from the output of _search or get _mapping API calls.
Here is the result of the _mapping call:
{
"[type]" : {
"properties" : {
"title" : {
"type" : "string"
}
"tags": {
"type" : "string"
}
}
}
}

Answer 1

Have a look at the example in the boost field documentation.
The boost field mapping (applied on the root object) allows to define
a boost field mapping where its content will control the boost level
of the document
The following mapping defines a field named _boost. If the _boost field itself exists within the JSON document indexed, its value will control the boost level of the indexed document.
{
"tweet" : {
"_boost" : {"name" : "_boost", "null_value" : 1.0}
}
}
Nothing special, the example just tells elasticsearch to consider the _boost field as it is and give a default 1.0 value to it when not present. But you are defining a boost for a specific document: that means that when the document matches a query, its score will be boosted according to the _boost field mapping that you applied to the root object. This doesn't have anything to do with boosting at a field level.
With your mapping you're saying that the content of the title field should be used as _boost, and you're giving a default _boost value of 2.0.
"_boost": {
"name": "title",
"null_value": 2.0
}
This doesn't make sense since the title contains text, and it's not either what you want I guess.
There are different ways to give more importance to a match on the title field.
As far as I understood from the documentation you can do it in your mapping like this:
{
"[type]" : {
"properties" : {
"title" : {
"type" : "string",
"boost" : 2.0
}
"tags": {
"type" : "string"
}
}
}
}
Quite honestly I haven't tried it and never used it before, but lucene does allow you to specify a boost per field at index time. The boost becomes part of the norms for that field and taken into account when there's a match on that specific field. So, this would be what you were looking for.
Anyway, I would personally do boosting at query time instead of index time, so that you don't need to modify your mapping and you can change the weight without reindexing. You can for example use a query string and search on different fields giving them different weights like this:
{
"query_string" : {
"fields" : ["title^2", "content"],
"query" : "this AND that OR thus"
}
}
You need to take into account that the query string query gets parsed and allows you to use the lucene query syntax.
Furthermore, you can combine different queries together using the bool query. You can express a boost for a match on title with a should clause containing for example a term query and a specific boost for it like this:
"should" : [
{
"term" : { "title" : "your query", "boost" : 2.0 }
}
]
You can use whatever query you want as should clause. If you go for the term query you need to remember it's not analyzed.