Custom PhraseQuery

[Blargy <zm...@hotmail.com>]

Is there anyway to override/change up the default PhraseQuery class that is
used... similar to how you can change out the Similarity class?

Let me explain what I am trying to do. I would like to override the TF is
calculated... always returning a max of 1 for phraseFreq. 

For example:
Query: "foo bar"
Doc1: "foo bar baz"
Doc2: "foo bar foo bar"

These two documents should be scored exactly the same. I accomplished the
above in the "normal" query use-case by using the SweetSpotSimilarity class.
There doesn't happen to be a SweetSpotPhraseQuery class is there?

Thanks
-- 
View this message in context: http://lucene.472066.n3.nabble.com/Custom-PhraseQuery-tp932414p932414.html
Sent from the Solr - User mailing list archive at Nabble.com.

Re: Custom PhraseQuery

[Chris Hostetter <ho...@fucit.org>]

: It sounds like all I need to do is actually override tf(float) in the
: SweetSpotSimilarity class to delegate to baselineTF just like tf(int) does.
: Is this correct?

you have to decide how you want to map the float->int (ie: round, 
truncate, etc...) but otherwise: yes that should work fine.



-Hoss

Re: Custom PhraseQuery

[Blargy <zm...@hotmail.com>]

Oh.. i didnt know about the different signatures to tf. Thanks for that
clarification.

It sounds like all I need to do is actually override tf(float) in the
SweetSpotSimilarity class to delegate to baselineTF just like tf(int) does.
Is this correct?

Thanks
-- 
View this message in context: http://lucene.472066.n3.nabble.com/Custom-PhraseQuery-tp932414p955257.html
Sent from the Solr - User mailing list archive at Nabble.com.

Re: Custom PhraseQuery

[Chris Hostetter <ho...@fucit.org>]

: Query: "foo bar"
: Doc1: "foo bar baz"
: Doc2: "foo bar foo bar"
: 
: These two documents should be scored exactly the same. I accomplished the
: above in the "normal" query use-case by using the SweetSpotSimilarity class.

You can change this by subclassing SweetSpotSimilarity (or any Similarity 
class) and overridding the tf(float) function.  

tf(int) is called for terms, while tf(float) is called for for phrases 
-- the float value is lower for phrases with a lot of slop, and higher for 
exact matches.

unfortunately, the input to tf(float) is lossy in accounting for docs 
htat match the phrase multiple times ... the value of "1.0f" 
might mean it mathes the phrase once exactly, or it might mean thta it 
matches many times in a sloppy manner.

in your case, it sounds like you just want it to return "1" for any input 
except "0.0f"



-Hoss