Boosting by facets with standard query

[ashokc <as...@qualcomm.com>]

I have a query that yields results binned in several facets. How can I boost
the results that fall in certain facets over the rest of them that do not
belong to those facets? I use the standard query format. Thank you
- ashok
-- 
View this message in context: http://www.nabble.com/Boosting-by-facets-with-standard-query-tp23084860p23084860.html
Sent from the Solr - User mailing list archive at Nabble.com.

Re: Boosting by facets with standard query

[Yonik Seeley <yo...@lucidimagination.com>]

On Fri, Apr 17, 2009 at 2:02 AM, ashokc <as...@qualcomm.com> wrote:
> q=(+(content:umts)+OR+(title:umts)^2+OR+(urltext:umts)^2)+AND+(doctype:white_papers)^2+AND+(filetype:pdf)^2
>
[...]
> What we need is for the white_papers & pdfs to be boosted, but if and only
> if such doucments are valid results to the search term in question.

Make the entire query part mandatory and the doctype boost optional.
Assuming that the default query operator is OR... this would look
like:
+(content:umts title:umts^2 urltext:umts^2) doctype:white_papers^2
filetype:pdf^2

You may also want to look at the dismax query parser, which can make
constructing queries like this a lot easier.

&defType=dismax
&q=umts
&qf=content title^2 urltext^2
&bq=doctype:white_papers^2 filetype:pdf^2


-Yonik
http://www.lucidimagination.com

Re: Boosting by facets with standard query

[ashokc <as...@qualcomm.com>]

Thanks for the tip. Looks like a neat idea. I have never used the sort
feature, so I have to create a new numeric key with values 1 or 2 - value 1
for white_papers/pdfs & 2 for others?

The problem also is that the facets I need to boost can vary by query. That
is, if the query term 'a', boost the facets 'facet1 & facet2'. If the query
term is 'b', then boost the facets 'facet4 & facet5'. Perhaps I can identify
the most freqently used boost order, and create as many fields as there are
orders. That would be the way right?
- ashok



Shalin Shekhar Mangar wrote:
> 
> On Fri, Apr 17, 2009 at 11:32 AM, ashokc <as...@qualcomm.com> wrote:
> 
>>
>> What we need is for the white_papers & pdfs to be boosted, but if and
>> only
>> if such doucments are valid results to the search term in question. How
>> would I write my above 'q' to accomplish that?
>>
> 
> Thanks for explaining in detail.
> 
> Basically, all you want to do is sort the results in the following order:
> 1. White papers
> 2. PDFs
> 3. Others
> 
> or maybe #1 and #2 are equivalent and can be intermingled.
> 
> Easiest way to do this is to index a new field whose values (when sorted)
> give you the desired order. Then you can simply sort on that field and
> score.
> 
> -- 
> Regards,
> Shalin Shekhar Mangar.
> 
> 

-- 
View this message in context: http://www.nabble.com/Boosting-by-facets-with-standard-query-tp23084860p23123288.html
Sent from the Solr - User mailing list archive at Nabble.com.

Re: Boosting by facets with standard query

[Shalin Shekhar Mangar <sh...@gmail.com>]

On Fri, Apr 17, 2009 at 11:32 AM, ashokc <as...@qualcomm.com> wrote:

>
> What we need is for the white_papers & pdfs to be boosted, but if and only
> if such doucments are valid results to the search term in question. How
> would I write my above 'q' to accomplish that?
>

Thanks for explaining in detail.

Basically, all you want to do is sort the results in the following order:
1. White papers
2. PDFs
3. Others

or maybe #1 and #2 are equivalent and can be intermingled.

Easiest way to do this is to index a new field whose values (when sorted)
give you the desired order. Then you can simply sort on that field and
score.

-- 
Regards,
Shalin Shekhar Mangar.

Re: Boosting by facets with standard query

[ashokc <as...@qualcomm.com>]

What you indicated here is for a different purpose, is it not? I already do
something similar with my 'q'. For example a sample query logged in
'catalina.out' looks like

webapp=/search path=/select
params={rows=15&start=0&q=(+(content:umts)+OR+(title:umts)^2+OR+(urltext:umts)^2)}

when the search term is "umts". I am looking for this term "umts" in the
fields  - (a) content, (b) title (boosted by a factor of 2) and (c) urltext
(boosted by a factor of 2). So the presense of the term "umts" in title or
url is weighed more than its presense in the regular content. So far so
good.

Now, I have other fields as well, like document type, file type etc... that
serve as facets to telescope down. Among the above set of search results, I
want to boost a specific document type 'white_papers' & a specific file type
"pdf". By boosting I mean that these white_paper & pdf documents should
float to the top of the heap in the search results, if such documents are at
all present in the search results.

So would I simply add the following to the above "q"?

q=(+(content:umts)+OR+(title:umts)^2+OR+(urltext:umts)^2)+AND+(doctype:white_papers)^2+AND+(filetype:pdf)^2

But wouldn't the above give 0 results if there are no white_papers & pdfs
(because of the AND)? If I use OR, then the meaning of the query is lost
altogether.

What we need is for the white_papers & pdfs to be boosted, but if and only
if such doucments are valid results to the search term in question. How
would I write my above 'q' to accomplish that?

Thanks

- ashok



Shalin Shekhar Mangar wrote:
> 
> On Fri, Apr 17, 2009 at 1:03 AM, ashokc <as...@qualcomm.com> wrote:
> 
>>
>> I have a query that yields results binned in several facets. How can I
>> boost
>> the results that fall in certain facets over the rest of them that do not
>> belong to those facets? I use the standard query format. Thank you
> 
> 
> I'm not sure what you mean by boosting by facet. Do you mean that you want
> to boost documents which match a term query?
> 
> If yes, you can use your_field_name:value^2.0 in the q parameter.
> -- 
> Regards,
> Shalin Shekhar Mangar.
> 
> 

-- 
View this message in context: http://www.nabble.com/Boosting-by-facets-with-standard-query-tp23084860p23091586.html
Sent from the Solr - User mailing list archive at Nabble.com.

Re: Boosting by facets with standard query

[Shalin Shekhar Mangar <sh...@gmail.com>]

On Fri, Apr 17, 2009 at 1:03 AM, ashokc <as...@qualcomm.com> wrote:

>
> I have a query that yields results binned in several facets. How can I
> boost
> the results that fall in certain facets over the rest of them that do not
> belong to those facets? I use the standard query format. Thank you


I'm not sure what you mean by boosting by facet. Do you mean that you want
to boost documents which match a term query?

If yes, you can use your_field_name:value^2.0 in the q parameter.
-- 
Regards,
Shalin Shekhar Mangar.