Improve suggestion tags for solr search query

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

Improve suggestion tags for solr search query

thakkar.aayush
Whenever a term is searched, we display the results and related tags. (We provide tags to each post using pre defined tags library) We would like to improve the tag suggestions to the user.

In the current implementation, related tags are shown based on facet counts for the search. But we intend to use combination of search relevance score (using lucene which uses vestor space model) with the search query and the facet count both.

We tried implementing the following :

Suggesting the tags to the user on the custom score that is uses facet count and search relevance score both. Mathematical model:

CustomScore= ((score - 1) * 0.3 + 1) * facetCount
score - Actual score of the faceted tags that lucene gives when searched against the search query.
facetCount - Facet count of the results extracted from the searches shown to the user.
0.3 - Boost factor

Any suggestion on this approach or better ways to get better tags based on relevancy and facet count?
Reply | Threaded
Open this post in threaded view
|

RE: Improve suggestion tags for solr search query

wmartinusa
Um, before I consider improving your algorithm, if I can,  is this a
homework problem? Are you making money off this search application?

Just need to be clear about such a give me a boost question, I think.

-----Original Message-----
From: thakkar.aayush [mailto:[hidden email]]
Sent: Monday, March 23, 2015 6:41 AM
To: [hidden email]
Subject: Improve suggestion tags for solr search query

Whenever a term is searched, we display the results and related tags. (We
provide tags to each post using pre defined tags library) We would like to
improve the tag suggestions to the user.

In the current implementation, related tags are shown based on facet counts
for the search. But we intend to use combination of search relevance score
(using lucene which uses vestor space model) with the search query and the
facet count both.

We tried implementing the following :

Suggesting the tags to the user on the custom score that is uses facet count
and search relevance score both. Mathematical model:

CustomScore= ((score - 1) * 0.3 + 1) * facetCount score - Actual score of
the faceted tags that lucene gives when searched against the search query.
facetCount - Facet count of the results extracted from the searches shown to
the user.
0.3 - Boost factor

Any suggestion on this approach or better ways to get better tags based on
relevancy and facet count?



--
View this message in context:
http://lucene.472066.n3.nabble.com/Improve-suggestion-tags-for-solr-search-q
uery-tp4194633.html
Sent from the Lucene - General mailing list archive at Nabble.com.

Reply | Threaded
Open this post in threaded view
|

RE: Improve suggestion tags for solr search query

thakkar.aayush
I am not using this as homework project. The suggestion is required for commercial application.
Any help is greatly regarded.

Thanks.
Reply | Threaded
Open this post in threaded view
|

RE: Improve suggestion tags for solr search query

wmartinusa
Thanks.
So the first thing I'm gonna do is build a model of how it might be working.
In Excel. Its' fun if you want to try.

More tonite.

w

-----Original Message-----
From: thakkar.aayush [mailto:[hidden email]]
Sent: Tuesday, March 24, 2015 2:25 AM
To: [hidden email]
Subject: RE: Improve suggestion tags for solr search query

I am not using this as homework project. The suggestion is required for
commercial application.
Any help is greatly regarded.

Thanks.



--
View this message in context:
http://lucene.472066.n3.nabble.com/Improve-suggestion-tags-for-solr-search-q
uery-tp4194633p4194842.html
Sent from the Lucene - General mailing list archive at Nabble.com.

Reply | Threaded
Open this post in threaded view
|

Re: Improve suggestion tags for solr search query

Malcolm Upayavira Holmes
In reply to this post by thakkar.aayush
If you want to do this in Solr, I’d suggest asking on the Solr user
list.

As I understand it, what you are asking will be hard to do. Scoring is
done, by Lucene, within the QueryComponent, which executes before the
FacetComponent has executed, so there is no knowledge of facet
values/counts at the time of scoring.

To enhance your score, you will need to use an existing Function Query,
or implement your own, as a ValueSourceParser.

Upayavira

On Mon, Mar 23, 2015, at 10:41 AM, thakkar.aayush wrote:

> Whenever a term is searched, we display the results and related tags. (We
> provide tags to each post using pre defined tags library) We would like
> to
> improve the tag suggestions to the user.
>
> In the current implementation, related tags are shown based on facet
> counts
> for the search. But we intend to use combination of search relevance
> score
> (using lucene which uses vestor space model) with the search query and
> the
> facet count both.
>
> We tried implementing the following :
>
> Suggesting the tags to the user on the custom score that is uses facet
> count
> and search relevance score both. Mathematical model:
>
> CustomScore= ((score - 1) * 0.3 + 1) * facetCount
> score - Actual score of the faceted tags that lucene gives when searched
> against the search query.
> facetCount - Facet count of the results extracted from the searches shown
> to
> the user.
> 0.3 - Boost factor
>
> Any suggestion on this approach or better ways to get better tags based
> on
> relevancy and facet count?
>
>
>
> --
> View this message in context:
> http://lucene.472066.n3.nabble.com/Improve-suggestion-tags-for-solr-search-query-tp4194633.html
> Sent from the Lucene - General mailing list archive at Nabble.com.