'Score keyword terms query on nested fields in elastichsearch 6.3
I have a set of keywords (skills in my example) and I would like to retrieve documents which match most of them. The documents should be sorted by how many of the keywords they match. The field i am searching into (skills) is of nested type. The index has the following mapping:
{
"mappings": {
"profiles": {
"properties": {
"id": {
"type": "keyword"
},
"skills": {
"type": "nested",
"properties": {
"level": {
"type": "float"
},
"name": {
"type": "keyword"
}
}
}
}
}
}
}
I tried both a terms query on the keyword field like:
{
"query": {
"nested": {
"path": "skills",
"query": {
"terms": {
"skills.name": [
"python",
"java"
]
}
}
}
}
}
And a boolean query
{
"query": {
"nested": {
"path": "skills",
"query": {
"bool": {
"should": [
{
"terms": {
"skills.name": [
"java"
]
}
},
{
"terms": {
"skills.name": [
"r"
]
}
}
]
}
}
}
}
}
For both queries the maximum score of the returned documents is 1. Thus both return documents that have ANY of the skills, but do not sort them such those with both skills are on top. The issues seems to be that skills is a nested field.
Solution 1:[1]
The second query works if each element of should is a nested query.
{
"query": {
"bool": {
"should": [
{
"nested": {
"path": "skills",
"query": {
"terms": {
"skills.name": [
"java"
]
}
}
}
},
{
"nested": {
"path": "skills",
"query": {
"terms": {
"skills.name": [
"r"
]
}
}
}
}
]
}
}
}
Sources
This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.
Source: Stack Overflow
| Solution | Source |
|---|---|
| Solution 1 | aymer |
