Reputation: 13
score for matching query_string documents
i'm currently working on a pretty annoying query i need from ES. my documents are nested docs, their index looks like something like this:
"mydocs" : {
"properties" : {
"doc" : {
"type" : "nested",
"properties" : {
"name" : {"type" : "string", "store" : "yes", "index" : "analyzed"},
"tagln" : {"type" : "string", "store" : "yes", "index" : "analyzed"},
"tags" : {"type" : "string", "store" : "yes", "index" : "analyzed"},
"featured" : {"type" : "integer", "store" : "yes", "index" : "not_analyzed"}
"blkd" : {"type" : "integer", "store" : "yes", "index" : "not_analyzed"},
... etc ...
}
i'm trying to boost the name, tagln and tags fields by a special score algo that adds the scores of featured*10000 + [is found in name]*1000 + [is found in tagln]*10 + [is found in tags]*10. my query is as follows:
{
"from" : 0,
"size" : 10,
"query" : {
"nested" : {
"query" : {
"filtered" : {
"query" : {
"bool" : {
"must" : [ {
"term" : {
"doc.blkd" : 0
}
} ],
"should" : [ {
"function_score" : {
"functions" : [ {
"field_value_factor" : {
"field" : "doc.featured",
"factor" : 10000.0
}
} ],
"score_mode" : "sum",
"boost_mode" : "sum"
}
}, {
"constant_score" : {
"filter" : {
"query_string" : {
"query" : "featured*",
"fields" : [ "doc.name^1000.0" ]
}
},
"boost" : 1000.0
}
}, {
"constant_score" : {
"filter" : {
"query_string" : {
"query" : "featured*",
"fields" : [ "doc.tags^10.0" ],
"boost" : 10.0
}
}
}
}, {
"constant_score" : {
"filter" : {
"query_string" : {
"query" : "featured*",
"fields" : [ "doc.tagln^10.0" ],
"boost" : 10.0
}
}
}
} ],
"minimum_should_match" : "0"
}
}
}
},
"path" : "doc",
"score_mode" : "sum"
}
},
"explain" : false,
"sort" : [ {
"_score" : { }
} ]
}
the score doesn't take the boosting into account as it should have, the score of the featured works as expected but the boost in the query_string doesn't work, docs with "aaa" in their names get a tiny score of 5 or 0. something while the featured=1 returns scores of 4000/6000/7500 etc..
first of all the score isn't 10000+ which is weird (might be due to many factors of the score) but the matching query string inside the name doesn't have any visible affect on the scores..
how can i solve this problem or atleast debug it better(to see how the score is being built)? tried changing explain to true but all i get is this pretty useless(or probably unreadable for me) explanation:
"_explanation": {
"value": 4000.0024,
"description": "sum of:",
"details": [
{
"value": 4000.0024,
"description": "Score based on child doc range from 387 to 387",
"details": []
},
{
"value": 0,
"description": "match on required clause, product of:",
"details": [
{
"value": 0,
"description": "# clause",
"details": []
},
{
"value": 0.0009999962,
"description": "-ConstantScore(_type:.percolator) #(+*:* -_type:__*), product of:",
"details": [
{
"value": 1,
"description": "boost",
"details": []
},
{
"value": 0.0009999962,
"description": "queryNorm",
"details": []
}
]
}
]
}
]
}
* edited *
thanks to keety i'm able to provide more info: after adding disable_coord-true and inner_hits explain-true i've tried "boosting" the query_string in any way i could.. the query is as follows:
{
"from" : 0,
"size" : 10,
"query" : {
"nested" : {
"query" : {
"filtered" : {
"query" : {
"bool" : {
"must" : [ {
"term" : {
"doc.blkd" : 0
}
} ],
"should" : [ {
"function_score" : {
"functions" : [ {
"field_value_factor" : {
"field" : "doc.featured",
"factor" : 10000.0
}
} ],
"score_mode" : "sum",
"boost_mode" : "sum"
}
}, {
"constant_score" : {
"filter" : {
"query_string" : {
"query" : "*featured*",
"fields" : [ "doc.name^1000.0" ]
}
},
"boost" : 1000.0
}
}, {
"query_string" : {
"query" : "*featured*",
"fields" : [ "doc.tags^100.0" ],
"boost" : 100.0
}
}, {
"constant_score" : {
"filter" : {
"query_string" : {
"query" : "*featured*",
"fields" : [ "doc.tagln^10.0" ],
"boost" : 10.0
}
}
}
} ],
"disable_coord" : true,
"minimum_should_match" : "0"
}
},
"filter" : {
"bool" : {
"should" : [ {
"query_string" : {
"query" : "*featured*",
"fields" : [ "doc.name^1000000.0", "doc.tags^10.0", "doc.tagln^10.0" ],
"boost" : 1000.0
}
} ],
"minimum_should_match" : "0"
}
}
}
},
"path" : "doc",
"score_mode" : "sum",
"inner_hits" : {
"explain" : "true"
}
}
},
"explain" : false,
"sort" : [ {
"_score" : { }
} ]
}
as you can see i've added the query_string to the filter and changed one of the query-shoulds to not be constant_score
the explanation of the doc now looks like this:
"max_score": 10001,
"hits": [
{
"_index": "myindex",
"_type": "mydocs",
"_id": "1111",
"_score": 10001,
"_ttl": 86158563,
"_source": {
"meta": {
"id": "1111",
"rev": "35-14602ccf5c3d429e0000000002000000",
"expiration": 0,
"flags": 33554432
},
"doc": {
"featured": 1,
"tagln": "hello location 1",
"blkd": 0,
"tags": [
"UsLocTaglinefeat"
],
"name": "hello US location featured"
}
},
"inner_hits": {
"doc": {
"hits": {
"total": 1,
"max_score": 10001,
"hits": [
{
"_shard": 1,
"_node": "YIXx2rrKR2O5q9519FIr_Q",
"_index": "myindex",
"_type": "mydocs",
"_id": "1111",
"_nested": {
"field": "doc",
"offset": 0
},
"_score": 10001,
"_source": {
"featured": 1,
"tagln": "hello location 1",
"blkd": 0,
"tags": [
"UsLocTaglinefeat"
],
"name": "hello US location featured"
},
"_explanation": {
"value": 10001,
"description": "sum of:",
"details": [
{
"value": 10001,
"description": "sum of:",
"details": [
{
"value": 0.0041682906,
"description": "weight(doc.blkd:`\b\u0000\u0000\u0000\u0000 in 0) [PerFieldSimilarity], result of:",
"details": [
{
"value": 0.0041682906,
"description": "score(doc=0,freq=1.0), product of:",
"details": [
{
"value": 0.0020365636,
"description": "queryWeight, product of:",
"details": [
{
"value": 2.0467274,
"description": "idf(docFreq=177, maxDocs=507)",
"details": []
},
{
"value": 0.0009950341,
"description": "queryNorm",
"details": []
}
]
},
{
"value": 2.0467274,
"description": "fieldWeight in 0, product of:",
"details": [
{
"value": 1,
"description": "tf(freq=1.0), with freq of:",
"details": [
{
"value": 1,
"description": "termFreq=1.0",
"details": []
}
]
},
{
"value": 2.0467274,
"description": "idf(docFreq=177, maxDocs=507)",
"details": []
},
{
"value": 1,
"description": "fieldNorm(doc=0)",
"details": []
}
]
}
]
}
]
},
{
"value": 10000.001,
"description": "sum of",
"details": [
{
"value": 0.0009950341,
"description": "*:*, product of:",
"details": [
{
"value": 1,
"description": "boost",
"details": []
},
{
"value": 0.0009950341,
"description": "queryNorm",
"details": []
}
]
},
{
"value": 10000,
"description": "min of:",
"details": [
{
"value": 10000,
"description": "field value function: none(doc['doc.featured'].value * factor=10000.0)",
"details": []
},
{
"value": 3.4028235e+38,
"description": "maxBoost",
"details": []
}
]
}
]
},
{
"value": 0.9950341,
"description": "ConstantScore(doc.name:*featured*), product of:",
"details": [
{
"value": 1000,
"description": "boost",
"details": []
},
{
"value": 0.0009950341,
"description": "queryNorm",
"details": []
}
]
}
]
},
{
"value": 0,
"description": "match on required clause, product of:",
"details": [
{
"value": 0,
"description": "# clause",
"details": []
},
{
"value": 0.0009950341,
"description": "((doc.name:*featured*)^1000000.0 | (doc.tags:*featured*)^10.0 | (doc.tagln:*featured*)^10.0), product of:",
"details": [
{
"value": 1,
"description": "boost",
"details": []
},
{
"value": 0.0009950341,
"description": "queryNorm",
"details": []
}
]
}
]
}
]
}
}
]
}
}
}
},
it seems like the only query_string affecting the score in any way is the one inside the filter but i can't seem to be able to boost it's score... any tips are welcome :) thanks
Upvotes: 1
Views: 2445
Answers (2)
Reputation: 13
"score_mode" : "sum",
"boost_mode" : "sum"
were my problems.. ES was normalizing the whole score but them and the result was weird because of that.
thanks to keety for the inner_hits explain.. it helped me a lot!
Upvotes: 0
Reputation: 17441
For the query in OP you would need to enable disable_coord in bool query to get the desired behaviour.
Also enabling inner_hits and setting explain:true in there would provide scoring details for nested documents. This feature is available in elasticsearch 1.5 onwards.
Example:
{
"query": {
"nested": {
"query": {
"filtered": {
"query": {
"bool": {
"disable_coord": "true",
"must": [
{
"term": {
"doc.blkd": 0
}
}
],
"should": [
{
"function_score": {
"functions": [
{
"field_value_factor": {
"field": "doc.featured",
"factor": 10000
}
}
],
"score_mode": "sum",
"boost_mode": "sum"
}
},
{
"constant_score": {
"filter": {
"query_string": {
"query": "featured*",
"fields": [
"doc.name^1000.0"
]
}
},
"boost": 1000
}
},
{
"constant_score": {
"filter": {
"query_string": {
"query": "featured*",
"fields": [
"doc.tags^10.0"
],
"boost": 10
}
}
}
},
{
"constant_score": {
"filter": {
"query_string": {
"query": "featured*",
"fields": [
"doc.tagln^10.0"
],
"boost": 10
}
}
}
}
],
"minimum_should_match": "0"
}
}
}
},
"path": "doc",
"score_mode": "sum",
"inner_hits" : {
"explain" : "true"
}
}
}
}
EDITED
Also it may be simpler to rewrite the query using function score as shown in below example.
{
"query": {
"nested": {
"query": {
"function_score": {
"functions": [
{
"field_value_factor": {
"field": "doc.featured",
"factor": 10000
}
},
{
"filter": {
"query_string": {
"query": "*featured*",
"fields": [
"doc.name^1000.0"
]
}
},
"weight": 1000
},
{
"filter": {
"query_string": {
"query": "*featured*",
"fields": [
"doc.tags^1000.0"
]
}
},
"weight": 100
},
{
"weight": 10,
"filter": {
"query_string": {
"query": "*featured*",
"fields": [
"doc.tagln^10.0"
]
}
}
}
],
"query": {
"term": {
"doc.blkd": 0
}
},
"score_mode": "sum",
"boost_mode": "sum"
}
},
"path": "doc",
"score_mode": "sum",
"inner_hits": {
"explain": "true"
}
}
}
}
Upvotes: 1
Related Questions
- Scoring in Elastic search
- Unsure of score in Elasticsearch
- ElasticSearch: given a document and a query, what is the relevance score?
- Elastic search score documents based on how close numerically a string is
- ElasticSearch match score
- ElasticSearch Function Score Query
- Score based on string in elasticsearch
- ElasticSearch match query score
- Elasticsearch: query accuracy
- Query match without score in elasticsearch