I am an Elastic Search newbie.
How can one make elastic search rank documents that more precisely match the input string?
For example, suppose we have the query
{
"query": {
"match": {
"name": "jones"
}
}
}
Suppose we have two documents:
Doc1: "name" : "jones"
Doc2: "name" : "jones jones jones jones jones"
I want Doc1 to be ranked more highly? It is a more precise match. How can I do this?
(Hopefully, in the most general possible way -- e.g. what if everywhere above 'jones' were replaced with 'fred jones')
Perhaps there are two approaches:
Maybe you can tell ES, "hey for this query a high term frequency should not be rewarded" (which seems to go against the core of ES, TF-IDF ...Because it very strongly wants to rewards a high TF (term frequency).
Maybe you can tell ES "prefer shorter matches over longer ones" (maybe using script_score???)
Surprised that I can't find answers to this question elsewhere. I must be missing something very fundamental.
Related
so I'm trying to get good search results out of an Elasticsearch installation.
But I run into problems when I'm trying to make a fuzzy search on some very simple data.
Somehow multiple (some of them partial) words are scored too low and only get scored higher, when more letters of the word are present in the search query.
Let me explain:
I have a simple index built with two simple documents.
{
"name": "Product with good qualities and awesome sound system"
},
{
"name": "Another Product that has better acustics than the other one"
}
Now I query the index with this parameters:
{
"query": {
"multi_match": {
"fields": ["name"],
"query": "product acust",
"fuzziness": "auto"
}
}
}
And the results look like this:
"hits": [
{
"_index": "test_products",
"_type": "_doc",
"_id": "1",
"_score": 0.19100355,
"_source": {
"name": "Product with good qualities and awesome sound system"
}
},
{
"_index": "test_products",
"_type": "_doc",
"_id": "2",
"_score": 0.17439455,
"_source": {
"name": "Another Product that has better acustics than the other one"
}
}
]
As you can see the product with the ID 2 is scored less than the other product even though it has possibly more similarity with the given query string than the other product because it has 1 full word match and 1 partial word match.
When the query would looke like "product acusti" the results would start to behave correctly.
I've already fiddled around with bool search but the results are identical.
Any ideas how I can get the wanted results back faster than having to have almost the whole second word typed in?
As far as I know, Elasticsearch does not do partial word matching by default, so the term acust is not matched in neither of your documents.
The reason you are getting a higher score in the first document is that your matched term, product, appears in a shorter sentence:
Product with good qualities and awesome sound system
But as for the second document, product appears in a longer sentence:
Another Product that has better acoustics than the other one
So your second document is getting a lower score because the ratio of your match term (product) to the number of terms in the sentence is lower.
In other words in has lower Field length normalization:
norm = 1/sqrt(numFieldTerms)
Now if you you want to be able to do partial prefix matching, you need to tokenize your term into ngrams, for example you can create the following ngrams for the term "acoustics":
"ac", "aco", "acou", "acous", "acoust", "acousti", "acoustic", "acoustics"
You have 2 options to achieve this, see the answer by Russ Cam on this question
use Analyze API
with an analyzer that will tokenize the field into tokens/terms from
which you would want to partial prefix match, and index this
collection as the input to the completion field. The Standard analyzer
may be a good one to start with...
Don't use the Completion Suggester here and instead set up your field (name) as a text datatype with
multi-fields
that include the different ways that name should be analyzed (or not
analyzed, with a keyword sub field for example). Spend some time with the Analyze API to build an analyzer that will
allow for partial prefix of terms anywhere in the name. As a start,
something like the Standard tokenizer, Lowercase token filter,
Edgengram token filter and possibly Stop token filter would get you
running...
You may also find this guide helpful.
I'm using elasticsearch to query data that originally was exported out of several relational databases that had a lot of redundencies. I now want to perform queries where I have a primary attribute and one or more secondary attributes that should match. I tried using a bool query with a must term and a should term, but that doesn't seem to work for my case, which may look like this:
Example:
I have a document with fullname and street name of a user and I want to search for similiar users in different indices. So the best match for my query should be the best match on fullname and best match on streetname field. But since the original data has a lot of redundencies and inconsistencies the field fullname (which I manually created out of fields name1, name2, name3) may contain the same name multiple times and it seems that elasticsearch ranks a double match in a must field higher than a match in a should attribute.
That means, I want to query for John Doe Back Street with the following sample data:
{
"fullname" : "John Doe John and Jane",
"street" : "Main Street"
}
{
"fullname" : "John Doe",
"street" : "Back Street"
}
Long story short, I want to query for a main attribute fullname - John Doe and secondary attribute street - Back Street and want the second document to be the best match and not the first because it contains John multiple times.
Manipulation of relevance in Elasticsearch is not the easiest part. Score calculation is based on three main parts:
Term frequency
Inverse document frequency
Field-length norm
Shortly:
the often the term occurs in field, the MORE relevant is
the often the term occurs in entire index, the LESS relevant is
the longer the term is, the MORE relevant is
I recommend you to read below materials:
What Is Relevance?
Theory Behind Relevance Scoring
Controlling Relevance and subpages
If in general, in your case, result of fullname is more important than from street you can boost importance of the first one. Below you have example code base on my working code:
{
"query": {
"multi_match": {
"query": "john doe",
"fields": [
"fullname^10",
"street"
]
}
}
}
In this example result from fullname is ten times (^10) much important than result from street. You can try to manipulate the boost or use other ways to control relevance but as I mentioned at the beginning - it is not the easiest way and everything depends on your particular situation. Mostly because of "inverse document frequency" part which considers terms from entire index - each next added document to index will probably change the score of the same search query.
I know that I did not answer directly but I hope to helped you to understand how this works.
I want to leverage ES fuzzy match for matching city names. For example, for these several inputs (Chicago, Chicago, IL, Chicago -IL, Chicago, USA, Chicago land, greater chicago area etc), I want to return one standard city name (Chicago,IL).
Simple fuzzy match works fine.
PUT /fuzzy_items/city/_mapping
{
"city": {
"properties": {
"name": {
"type": "string",
"analyzer": "simple"
}
}
}
}
However, I've several documents (ex. 10,000) which needs to be normalized. Other operations have bulk API.. ie. If I need to index 10,000 docs, bulk API is handy to achieve this purpose in fewer calls.
Is there a similar feature for fuzzy match? meaning, can I send array of inputs ["chicago,USA","Minneapolis, MN"] and expect a fuzzy matched response array ["Chicago,IL","Minneapolis,MN"]?
The problem I'm trying to overcome is how 'not' to make 1000s of calls to fuzzy match and achieve the same with fewer calls.
Or
Is there a different way of achieving this, may be using elastic scripting ?
Appreciate any suggestions.
My use case is like this:
for a query iphone charger, I am getting higher relevance for results, having name, iphone charger coupons than with name iphone charger, possibly because of better match in description and other fields. Boosting name field isn't helping much unless I skew the importance drastically. what I really need is tf/idf boost within name field
to quote elasticsearch blog:
the frequency of a term in a field is offset by the length of the field. However, the practical scoring function treats all fields in the same way. It will treat all title fields (because they are short) as more important than all body fields (because they are long).
I need to boost this more important value for a particular field. Can we do this with function score or any other way?
A one term difference in length is not much of a difference to the scoring algorithm (and, in fact, can vanish entirely due to imprecision on the length norm). If there are hits on other fields, you have a lot of scoring elements to fight against.
A dis_max would probably be a reasonable approach to this. Instead of all the additive scores and coords and such you are trying to overcome, it will simply select the score of the best matching subquery. If you boost the query against title, you can ensure matches there are strongly preferred.
You can then assign a "tie_breaker", so that the score against the description subquery is factored in only when "title" scores are tied.
{
"dis_max" : {
"tie_breaker" : 0.2,
"queries" : [
{
"terms" : {
"age" : ["iphone", "charger"],
"boost" : 10
}
},
{
"terms" : {
"description" : ["iphone", "charger"]
}
}
]
}
}
Another approach to this sort of thing, if you absolutely know when you have an exact match against the entire field, is to separately index an untokenized version of that field, and query that field as well. Any match against the untokenized version of the field will be an exact match again the entire field contents. This would prevent you needing to relying on the length norm to make that determination.
I am new to elasticsearch. In elasticsearch we can use the term boost in almost all queries. I understand it's used for modify score of documents. But i can't find actual use of it. My query is if i use boost values in some queries, will it affect final score of search or the boost rank of docs in index itself.
And what is main difference between boost at index and boost at querying..
Thanks in Advance..!
Query time boost allows you to give more weight to one query than to another. For instance, let's say you are querying the title and body fields for "Quick Brown Fox", you could write it as:
{
"query": {
"bool": {
"should": [
{
"match": {
"title": "Quick Brown Fox"
}
},
{
"match": {
"body": "Quick Brown Fox"
}
}
]
}
}
}
But you decide that you want the title field to be more important than the body field, which means you need to boost the query on the title field by (eg) 2:
{
"query": {
"bool": {
"should": [
{
"match": {
"title": {
"query": "Quick Brown Fox",
"boost": 2
}
}
},
{
"match": {
"body": "Quick Brown Fox"
}
}
]
}
}
}
(Note how the structure of the match clause changed to accommodate the boost parameter).
The boost value of 2 doesn't double the _score exactly - the scores go through a normalization process. So you should think of boost as make this query clause relatively more important than the other query clauses.
My doubt is if i use boost values in some queries. will it affect final score of search
Yes it does, but you shouldn't rely on the actual value of _score anyway. Its only purpose is to allow Elasticsearch to decide which documents are most relevant to this query. If the query changes, the scores change.
Re index time boosting: don't use it. It's inflexible and error prone.
Boost at query time won't modify your index. It only applies boost factor on fields when searching.
I prefer boost at query time as it's more flexible. If you need to change your boost rules and you had set it at index time, you will probably need to reindex.
Use cases of boosting : Suppose you are building a e-commerce web app, and your product data is in elastic search. Whenever a customer uses search bar you query elastic search and displays the result in web app.
Elastic search keeps relevance score for every document and returns the result in sorted order of the relevance score.
Now let's assume a user searches for "samsung phones", then should your web app just show samsung phones -> Answer is NO.
Your web app should show other phones as well (as user may like those as well) but first show samsung phones (as he/she is looking for those) and then show other phones as well.
So question is how do you query where samsung phones comes up in result ? -> Answer is relevance score.
Let say you hit query like for all mobile phones and samsung phone and the keep high relevance score of samsung phones,
Then result will contain first samsung phones and then other phones.