Login | Register For Free | Help
Search for: (Advanced)

Mailing List Archive: Lucene: General

Improving relevance by document date

 

 

Lucene general RSS feed   Index | Next | Previous | View Threaded


amaskaev at gmail

Jan 29, 2010, 9:56 AM

Post #1 of 3 (760 views)
Permalink
Improving relevance by document date

Hi all,

I have the following problem. My lucene index contains documents with fields
Text and CreatedDate.
I'd like to have document search results sorted by relevance but also wants
to show newest documents first within reasonable limits.

I can create sorting by relevance and then by date, but this approach
provide too rough results, because even the oldest documents with higher
rank will come first. I'd like to have more elegant solution.

It seems that i should include document date to scoring formula to get more
rank to the newer documents. Probably I can set document boost based on
date, so newest documents will have bigger boost value. But i'm not sure is
this variant correct.

Maybe someone can suggest more correct approach?


--
View this message in context: http://old.nabble.com/Improving-relevance-by-document-date-tp27375856p27375856.html
Sent from the Lucene - General mailing list archive at Nabble.com.


ted.dunning at gmail

Jan 29, 2010, 11:42 AM

Post #2 of 3 (726 views)
Permalink
Re: Improving relevance by document date [In reply to]

Don't set a document boost. Just add a date range to all queries.

On Fri, Jan 29, 2010 at 9:56 AM, Almas1997 <amaskaev [at] gmail> wrote:

>
> Hi all,
>
> I have the following problem. My lucene index contains documents with
> fields
> Text and CreatedDate.
> I'd like to have document search results sorted by relevance but also wants
> to show newest documents first within reasonable limits.
>
> I can create sorting by relevance and then by date, but this approach
> provide too rough results, because even the oldest documents with higher
> rank will come first. I'd like to have more elegant solution.
>
> It seems that i should include document date to scoring formula to get more
> rank to the newer documents. Probably I can set document boost based on
> date, so newest documents will have bigger boost value. But i'm not sure is
> this variant correct.
>
> Maybe someone can suggest more correct approach?
>
>
> --
> View this message in context:
> http://old.nabble.com/Improving-relevance-by-document-date-tp27375856p27375856.html
> Sent from the Lucene - General mailing list archive at Nabble.com.
>
>


--
Ted Dunning, CTO
DeepDyve


amaskaev at gmail

Feb 1, 2010, 12:37 AM

Post #3 of 3 (722 views)
Permalink
Re: Improving relevance by document date [In reply to]

Thank you.

However just found more useful approach using CustomScoreQuery.


Ted Dunning wrote:
>
> Don't set a document boost. Just add a date range to all queries.
>
> On Fri, Jan 29, 2010 at 9:56 AM, Almas1997 <amaskaev [at] gmail> wrote:
>
>>
>> Hi all,
>>
>> I have the following problem. My lucene index contains documents with
>> fields
>> Text and CreatedDate.
>> I'd like to have document search results sorted by relevance but also
>> wants
>> to show newest documents first within reasonable limits.
>>
>> I can create sorting by relevance and then by date, but this approach
>> provide too rough results, because even the oldest documents with higher
>> rank will come first. I'd like to have more elegant solution.
>>
>> It seems that i should include document date to scoring formula to get
>> more
>> rank to the newer documents. Probably I can set document boost based on
>> date, so newest documents will have bigger boost value. But i'm not sure
>> is
>> this variant correct.
>>
>> Maybe someone can suggest more correct approach?
>>
>>
>> --
>> View this message in context:
>> http://old.nabble.com/Improving-relevance-by-document-date-tp27375856p27375856.html
>> Sent from the Lucene - General mailing list archive at Nabble.com.
>>
>>
>
>
> --
> Ted Dunning, CTO
> DeepDyve
>
>

--
View this message in context: http://old.nabble.com/Improving-relevance-by-document-date-tp27375856p27401068.html
Sent from the Lucene - General mailing list archive at Nabble.com.

Lucene general RSS feed   Index | Next | Previous | View Threaded
 
 


Interested in having your list archived? Contact Gossamer Threads
 
  Web Applications & Managed Hosting Powered by Gossamer Threads Inc.