lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From aurelien.mazo...@francelabs.com
Subject Re: Use SloppyPhraseScorer in SpanNearQuery
Date Mon, 14 Sep 2015 07:58:29 GMT
 

Hi Alan, 

Thank you for your answer: it is exactly what I need. 

Regards, 

Aurélien 

On 12.09.2015 11:21, Alan Woodward wrote: 

> If you're using lucene 5.3, you could create a modified version of PayloadScoreQuery
that just uses positions rather than payloads to modify the score. 
> 
> Alan Woodward
> www.flax.co.uk [1]
> 
> On 11 Sep 2015, at 17:40, aurelien.mazoyer@francelabs.com wrote: 
> 
> Perfect, thank you :-) 
> 
> I was also wondering if I can take the absolute position of a term in the scoring calculation
in my SpanNearQuery but I am not sure I can achieve this efficiently with a custom implementation
of Similarity. 
> 
> I am thinking of two ways to achieve this (please tell me if I am wrong...) : 
> 
> - Duplicate the term position in payload and implement the computePayload method in Similirity
class in order to make it take the value into account for scoring (I don't like this solution
:-) ) 
> 
> - Extends the SpanScorer to take the term position into account for the scoring calculation

> 
> Do you think that one of these approach is valid or do you have another idea? 
> 
> Thank you, 
> 
> Regards 
> 
> Aurélien MAZOYER 
> 
> On 11.09.2015 09:40, Alan Woodward wrote: Yes, that's correct. 
> 
> Alan Woodward
> www.flax.co.uk [2]
> 
> On 10 Sep 2015, at 17:11, aurelien.mazoyer@francelabs.com wrote: Hi,
> 
> Yes, thanks.
> If I want to customize this behavior (remove it or make it more or less important), is
it correct to override the Similarity class to implement a new computeSlopFactor method?
> 
> Thank you for your answer.
> 
> Regards,
> 
> Aurélien MAZOYER
> 
> On 10.09.2015 12:37, Alan Woodward wrote:
> Hi,
> SpanNearQuery will also take into account the 'width' of the match, so
> that terms that are closer together will score more highly. Is that
> what you're looking for?
> Alan Woodward
> www.flax.co.uk [2]
> On 10 Sep 2015, at 10:43, aurelien.mazoyer@francelabs.com wrote:
> Hi all,
> SpanNearQuery and SloppyPhraseQuery can have similar behavior in some use cases. On the
one hand, I have to distinguish between ordered and unordered terms (that's enabled with SpanNearQuery)
but on the other hand, I want to use the SloppyPhraseScorer, that takes the relative positions
of the terms into account for the score calculation. As far I understood, this is possible
with SloppyPhraseQuery but not with SpanNearQuery. Do you think there is there a way to combine
these two features to have the best of both worlds?
> Thanks!
> Aurélien
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
> For additional commands, e-mail: java-user-help@lucene.apache.org

 

Links:
------
[1] http://www.flax.co.uk
[2] http://www.flax.co.uk/

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message