elasticsearch conditional Boost Index time and Query time - elasticsearch

Question

I have twitter_user_index
I want to boost tweets score in index time
If user.verified:
boost: 10
elif user.follower_count in range (1, 100):
boost: 1
elif user.follower_count in range (101, 200):
boost: 2
How do I achieve this kind of boosting in index time or in query time (both answers welcome)

Answer 1

I have taken an index with below mapping. "user" I have taken as "object type", if your type is "nested" then you need to use nested query, rest structure will be same.
Mappings:
PUT twitter_user_index
{
"mappings": {
"properties": {
"user":{
"type": "object",
"properties": {
"name":{
"type":"text",
"fields":{
"keyword":{
"type":"keyword"
}
}
},
"verified":{
"type":"boolean"
},
"follower_count":{
"type": "integer"
}
}
}
}
}
}
Data: I have taken three records. 2 users with verified account having less than 100 and more than 100 followers respectively and one user with non verified account having more than 100 followers
"hits" : [
{
"_index" : "twitter_user_index",
"_type" : "_doc",
"_id" : "gwiu-HEBZgLhu13ZIerO",
"_score" : 1.0,
"_source" : {
"user" : {
"name" : "abc",
"verified" : true,
"follower_count" : 90
}
}
},
{
"_index" : "twitter_user_index",
"_type" : "_doc",
"_id" : "hAiu-HEBZgLhu13Za-qL",
"_score" : 1.0,
"_source" : {
"user" : {
"name" : "efg",
"verified" : true,
"follower_count" : 120
}
}
},
{
"_index" : "twitter_user_index",
"_type" : "_doc",
"_id" : "hQiu-HEBZgLhu13ZhOrr",
"_score" : 1.0,
"_source" : {
"user" : {
"name" : "xyz",
"verified" : false,
"follower_count" : 120
}
}
}
]
Query: I have used a should clause which will not filter out any document, one score matching documents higher. Each clause has its boost to increase relevance.
GET twitter_user_index/_search
{
"query": {
"bool": {
"should": [
{
"term": {
"user.verified": {
"value": true,
"boost": 10
}
}
},
{
"range": {
"user.follower_count": {
"gte": 1,
"lte": 100,
"boost": 1
}
}
},
{
"range": {
"user.follower_count": {
"gte": 101,
"lte": 200,
"boost": 2
}
}
}
]
}
}
}
Result: In output of query, verified accounts are scored higher than non verified account and accounts with followers in given range are scored higher.
"hits" : [
{
"_index" : "twitter_user_index",
"_type" : "_doc",
"_id" : "hAiu-HEBZgLhu13Za-qL",
"_score" : 6.700036,
"_source" : {
"user" : {
"name" : "efg",
"verified" : true,
"follower_count" : 120
}
}
},
{
"_index" : "twitter_user_index",
"_type" : "_doc",
"_id" : "gwiu-HEBZgLhu13ZIerO",
"_score" : 5.700036,
"_source" : {
"user" : {
"name" : "abc",
"verified" : true,
"follower_count" : 90
}
}
},
{
"_index" : "twitter_user_index",
"_type" : "_doc",
"_id" : "hQiu-HEBZgLhu13ZhOrr",
"_score" : 2.0,
"_source" : {
"user" : {
"name" : "xyz",
"verified" : false,
"follower_count" : 120
}
}
}
]
EDIT1: Elasticsearch DSL if else blocks cannot be added. We can add additonal should clause that verified:false and range between 1 and 100 then select document applying boost.
{
"query": {
"bool": {
"should": [
{
"term": {
"user.verified": {
"value": true,
"boost": 10
}
}
},
{
"bool": {
"should": [
{
"term": {
"user.verified": {
"value": false
}
}
},
{
"range": {
"user.follower_count": {
"gte": 1,
"lte": 100,
"boost": 1
}
}
}
]
}
},
{
"bool": {
"should": [
{
"term": {
"user.verified": {
"value": false
}
}
},
{
"range": {
"user.follower_count": {
"gte": 101,
"lte": 200,
"boost": 2
}
}
}
]
}
}
]
}
}
}
Result
"hits" : [
{
"_index" : "twitter_user_index",
"_type" : "_doc",
"_id" : "hAiu-HEBZgLhu13Za-qL",
"_score" : 5.566749,
"_source" : {
"user" : {
"name" : "efg",
"verified" : true,
"follower_count" : 120
}
}
},
{
"_index" : "twitter_user_index",
"_type" : "_doc",
"_id" : "hgi4-HEBZgLhu13ZW-rY",
"_score" : 5.566749,
"_source" : {
"user" : {
"name" : "xyz1",
"verified" : true,
"follower_count" : 150
}
}
},
{
"_index" : "twitter_user_index",
"_type" : "_doc",
"_id" : "gwiu-HEBZgLhu13ZIerO",
"_score" : 4.566749,
"_source" : {
"user" : {
"name" : "abc",
"verified" : true,
"follower_count" : 90
}
}
},
{
"_index" : "twitter_user_index",
"_type" : "_doc",
"_id" : "hQiu-HEBZgLhu13ZhOrr",
"_score" : 4.4079456,
"_source" : {
"user" : {
"name" : "xyz",
"verified" : false,
"follower_count" : 120
}
}
}
]
EDIT3
Optimized version using filter clause. Filter doesn't calculate score so is faster
{
"query": {
"bool": {
"should": [
{
"term": {
"user.verified": {
"value": true,
"boost": 10
}
}
},
{
"bool": {
"filter": [
{
"term": {
"user.verified": {
"value": false
}
}
}],
"must":[{
"range": {
"user.follower_count": {
"gte": 1,
"lte": 100,
"boost": 1
}
}
}
]
}
},
{
"bool": {
"filter": [
{
"term": {
"user.verified": {
"value": false
}
}
}],
"must":[
{
"range": {
"user.follower_count": {
"gte": 101,
"lte": 200,
"boost": 2
}
}
}
]
}
}
]
}
}
}