Need advise on indexing/search strategy

Elastic Stack Elasticsearch
1

Hi

I have list of products, where each product will have category, color,
designer and description.

The problem i am trying to solve is search relevancy.

I started by indexing everything as-is, but results where not satisfying -
since some products had words used for category name put into description
or designer names, results were score unfairly and relevancy was low.

Next thing i tried was to create stopwords. So basically any word used as
category name was put into other field stopfilter,
and the same for colors. This greatly improved relevancy. A product with
color/category name in description or designer will not bias search result.
Search result for most of test cases are very good, however one problem
remained: ambiguity.

let's say we products in category carpets, some will have color red and
there is designer named "red johh". When searching
for red carpets, i'd like to score carpets in color red higher then
anything else. However "red john carpets" should find
carpets designed by "red john", and not just anything that's red, john or
carpet. Stopwords filter solution removes red from designer field,
so it can't be found this way. What is the best way to solve this problem?

Thanks for any help.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

2

If you index products and categories, why don't you just search in
different fields?

For boosting query terms, you could use "boost"
http://www.elasticsearch.org/guide/reference/query-dsl/term-query/ to push
documents with category carpet to the top if somebody entered "red".

To disambiguate "red john" from "red" you need entity recognition before
sending the query to ES.

Jörg

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

3

Jorg is right. First you ring a bell, then you blow a whistle and then you
look for some other tool to ask to become a bell or whistle. In the
meantime, you know nothing about your data or relevancy models. Good luck,

On Saturday, August 24, 2013 7:08:56 AM UTC-4, Maciej Dziardziel wrote:

Hi

I have list of products, where each product will have category, color,
designer and description.

The problem i am trying to solve is search relevancy.

I started by indexing everything as-is, but results where not satisfying -
since some products had words used for category name put into description
or designer names, results were score unfairly and relevancy was low.

Next thing i tried was to create stopwords. So basically any word used as
category name was put into other field stopfilter,
and the same for colors. This greatly improved relevancy. A product with
color/category name in description or designer will not bias search result.
Search result for most of test cases are very good, however one problem
remained: ambiguity.

let's say we products in category carpets, some will have color red and
there is designer named "red johh". When searching
for red carpets, i'd like to score carpets in color red higher then
anything else. However "red john carpets" should find
carpets designed by "red john", and not just anything that's red, john or
carpet. Stopwords filter solution removes red from designer field,
so it can't be found this way. What is the best way to solve this problem?

Thanks for any help.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.