You are viewing a plain text version of this content. The canonical link for it is here.
Posted to solr-user@lucene.apache.org by Chris <ch...@gmail.com> on 2013/11/10 06:47:14 UTC
Query Relevancy tuning...
Hi Gurus,
I have a relevancy ranking questrion -
1. I have fields - title, domain, domainrank in index.
2. I am looking to maybe load a txt file of prefered domains at solr
startup & boost documents from those domains if keyword matches text in
title or domain (if it exactly matches the domain, it should rank higher,
than if it were a semi match)
3. Also, i would like to have 2-3 results per domain per page.(at the max)
4. Also, is it possible to do intersection - if all 4 words(say) matches it
should rank higher than maybe 3 word match & so on..
I would like this to be as fast as possible, so kindly suggest an optimal
way of doing this.
a few things that were tried....
<str name="defType">edismax</str>
<str name="qf">
fulltxt^0.5 title^2.0 domain^3 urlKeywords^1.5 anchorText^2.0
h1Keywords^1.5
</str>
<str name="df">text</str>
<str name="mm">100%</str>
<str name="q.alt">*:*</str>
<str name="rows">10</str>
<str name="fl">*,score</str>
Re: Query Relevancy tuning...
Posted by Upayavira <uv...@odoko.co.uk>.
Some pointers for your research:
Look at the ExternalFileField, which will get you your boosting if
fields match specific values [1].
You can get 2 or 3 values per domain by doing field collapsing.
By default, the scoring that Solr uses includes a 'co-ordination factor'
which adjusts the score according to how many of your query terms
matched a document, so you shouldn't need to worry about all of this.
Upayavira
[1]
https://cwiki.apache.org/confluence/display/solr/Working+with+External+Files+and+Processes
[2] http://wiki.apache.org/solr/FieldCollapsing
On Sun, Nov 10, 2013, at 05:47 AM, Chris wrote:
> Hi Gurus,
>
> I have a relevancy ranking questrion -
>
> 1. I have fields - title, domain, domainrank in index.
> 2. I am looking to maybe load a txt file of prefered domains at solr
> startup & boost documents from those domains if keyword matches text in
> title or domain (if it exactly matches the domain, it should rank higher,
> than if it were a semi match)
> 3. Also, i would like to have 2-3 results per domain per page.(at the
> max)
> 4. Also, is it possible to do intersection - if all 4 words(say) matches
> it
> should rank higher than maybe 3 word match & so on..
>
> I would like this to be as fast as possible, so kindly suggest an optimal
> way of doing this.
>
> a few things that were tried....
>
> <str name="defType">edismax</str>
> <str name="qf">
> fulltxt^0.5 title^2.0 domain^3 urlKeywords^1.5 anchorText^2.0
> h1Keywords^1.5
> </str>
> <str name="df">text</str>
> <str name="mm">100%</str>
> <str name="q.alt">*:*</str>
> <str name="rows">10</str>
> <str name="fl">*,score</str>