NutchQuery adding non required Terms

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

NutchQuery adding non required Terms

Stefan Groschupf-2
Hi,
I would love to build a nutch Query object via API and not using the  
Queryparser.
In my case I need the complete set of boolean operators in the query,  
so required (AND) and non required (OR)  terms and prohibited (NOT).
I notice that in general this would be possible to add a clause in  
the Query object, since the BasicQuery filter just copies the  
parameter isRequired and isProhibited.
However the Clauses arraylist is private and there is not method in  
the nutch query object that allows to add custom terms or clauses  
with isRequired and isProhibited.

Did I miss something in general to be able to support non required  
terms in nutch?
Would people agree to add a little method that allows to adding terms  
with these parameters?


Thanks for any comments.
Stefan
Reply | Threaded
Open this post in threaded view
|

Re: NutchQuery adding non required Terms

Doug Cutting-2
Stefan Groschupf wrote:
> Did I miss something in general to be able to support non required  
> terms in nutch?

I left OR and nesting out of the API to simplify what query filters have
to process.  Nutch's query features are approximately what Google
supported for its first three years.  (Google did not add OR until 2000,
I think.)

If we permit optional clauses then we need to make sure that each query
filter can handle them correctly.

For example, the query "+A +B" is translated by query-basic into
something like:

+(title:a OR content:a OR anchors:a OR url:a OR host:a)
+(title:b OR content:b OR anchors:b OR url:b OR host:b)
title:"a b"~999
content:"a b"~999
anchors:"a b"~999
url:"a b"~999
host:"a b"~999

The query "+A B" (where B is optional) should remove the plus in the
second line above.  So it should not be too hard to change query-basic
to be able to handle optional terms in the default field.  Perhaps
that's the only query filter that would need to be updated.  And it
looks like LuceneQueryOptimizer already checks that filterized clauses
are required.

It would be good to have some unit tests for query filtering.

Doug
Reply | Threaded
Open this post in threaded view
|

Re: NutchQuery adding non required Terms

Stefan Groschupf-2
Thanks for the hint.
I would love to add non required terms and nesting to the Query  
object API, I will provide also some unit tests, but since I'm not a  
javacc geek it will only extend the java api not the query parser.
Would such a extension  be welcome?

Stefan

Am 12.01.2006 um 18:29 schrieb Doug Cutting:

> Stefan Groschupf wrote:
>> Did I miss something in general to be able to support non  
>> required  terms in nutch?
>
> I left OR and nesting out of the API to simplify what query filters  
> have to process.  Nutch's query features are approximately what  
> Google supported for its first three years.  (Google did not add OR  
> until 2000, I think.)
>
> If we permit optional clauses then we need to make sure that each  
> query filter can handle them correctly.
>
> For example, the query "+A +B" is translated by query-basic into  
> something like:
>
> +(title:a OR content:a OR anchors:a OR url:a OR host:a)
> +(title:b OR content:b OR anchors:b OR url:b OR host:b)
> title:"a b"~999
> content:"a b"~999
> anchors:"a b"~999
> url:"a b"~999
> host:"a b"~999
>
> The query "+A B" (where B is optional) should remove the plus in  
> the second line above.  So it should not be too hard to change  
> query-basic to be able to handle optional terms in the default  
> field.  Perhaps that's the only query filter that would need to be  
> updated.  And it looks like LuceneQueryOptimizer already checks  
> that filterized clauses are required.
>
> It would be good to have some unit tests for query filtering.
>
> Doug
>

---------------------------------------------------------------
company:        http://www.media-style.com
forum:        http://www.text-mining.org
blog:            http://www.find23.net


Reply | Threaded
Open this post in threaded view
|

Re: NutchQuery adding non required Terms

Doug Cutting-2
Stefan Groschupf wrote:
> I would love to add non required terms and nesting to the Query  object
> API, I will provide also some unit tests, but since I'm not a  javacc
> geek it will only extend the java api not the query parser.
> Would such a extension  be welcome?

I think we should start with just adding non-required terms, and leave
nesting as a subsequent step.

I also agree that we can leave this out of the query parser as a start.

Doug