Good to know you found a way through Tom :)

> On 8 Apr 2015, at 11:12 pm, Tom Davies <[email protected]> wrote:
> 
> Hi Walter,
> 
> Thanks for the idea.  However, in my case the volume of data requires I take 
> advantage of Sphinx if possible (the site is http://teenormous.com).
> 
> The previous ranker did not add to the weight score when duplicates were 
> within the same search field.  So, my previous approach to limit keyword 
> stuffing was to put all the fields I wanted to have the same impact on the 
> results into one combined search_data field.  However, this is no longer 
> working as I am seeing duplicate keywords cause a slightly higher weight 
> score.
> 
> I just found the answer as I was typing this.  The solution (for me) to 
> ignore duplicate works is to use the :proximity ranker, rather than the 
> default :proximity_bm25.  For example:
> 
> Article.search "pancakes", :ranker => :proximity
> 
> The clue came from this thread: 
> http://sphinxsearch.com/forum/view.html?id=3675.  Specifically this portion 
> talking about the weight calculation for the default :proximity_bm25 ranker: 
> 
> " the last three digits, are the 3 most significant decimal places from the 
> BM25 ranking algorithm (whose aim is to rank rare words higher, or many 
> instances of a word in a single document higher)."
> 
> Thanks,
> 
> Tom
> 
> On Wednesday, April 8, 2015 at 8:06:24 AM UTC-4, Walter Davis wrote:
> I don't know how to do this in Sphinx per se, but what about adding a 
> keywords column to your table, and only index that. Then you could control 
> what makes it in there with some logic (split and uniq and join).
> 
> Walter
> 
> On Apr 7, 2015, at 5:58 PM, Tom Davies <[email protected] <javascript:>> 
> wrote:
> 
>> I just recently upgraded from ThinkingSphinx from 2.0.14 to 3.1.3 and Sphinx 
>> from 2.0.4 to 2.2.8 and for the most part everything is working well.
>> 
>> However, I noticed a slight change to the default weight calculation based 
>> on the new ranker.  The new default ranker appears to be adding additional 
>> weight for repeated phrases within the same field.
>> 
>> Is there an easy way to only apply a weight for one occurrence in each field?
>> 
>> For example, if I have two t-shirts:
>> 
>> Tee 1:  
>> 
>> name: "Star Wars"
>> 
>> Tee 2:
>> 
>> name: "Star Wars Star Wars"
>> 
>> I would like them to end up with the same weight.  In my case, I have 
>> t-shirt data from a variety of merchants and many of them are prone to 
>> stuffing keywords over and over.  This is causing them to get an extra bump 
>> as a result.
>> 
>> Thanks!
>> Tom
>> 
>> 
>> -- 
>> You received this message because you are subscribed to the Google Groups 
>> "Thinking Sphinx" group.
>> To unsubscribe from this group and stop receiving emails from it, send an 
>> email to [email protected] <javascript:>.
>> To post to this group, send email to [email protected] 
>> <javascript:>.
>> Visit this group at http://groups.google.com/group/thinking-sphinx 
>> <http://groups.google.com/group/thinking-sphinx>.
>> For more options, visit https://groups.google.com/d/optout 
>> <https://groups.google.com/d/optout>.
> 
> 
> -- 
> You received this message because you are subscribed to the Google Groups 
> "Thinking Sphinx" group.
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to [email protected] 
> <mailto:[email protected]>.
> To post to this group, send email to [email protected] 
> <mailto:[email protected]>.
> Visit this group at http://groups.google.com/group/thinking-sphinx 
> <http://groups.google.com/group/thinking-sphinx>.
> For more options, visit https://groups.google.com/d/optout 
> <https://groups.google.com/d/optout>.

-- 
You received this message because you are subscribed to the Google Groups 
"Thinking Sphinx" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/thinking-sphinx.
For more options, visit https://groups.google.com/d/optout.

Reply via email to