You are viewing a plain text version of this content. The canonical link for it is here.
Posted to solr-user@lucene.apache.org by Harsha JSN <ha...@gmail.com> on 2016/06/24 05:21:41 UTC
Using n-grams vs AnalyzingInfixLookupFactory for suggestions in solr
Hi,
I have some doubts regarding usage of AnalyzingInfixLookupFactory as
lookup implementation for suggestions.
1.) AnalyzingInfixLookupFactory constructs n-grams for the suggestion field
while building suggestions index. If the main index which is used for
search is already having n-grams for this field, is it still preferred to
choose AnalyzingInfixLookupFactory or can we directly build suggestions
from the main index?
2.) Also, AnalyzingInfixLookupFactory returns duplicate records if the
suggestion field has same value in multiple documents. Instead if i search
suggestions from main index (n-grams) i can eliminate the duplicates by
grouping the results. But grouping can be a complex operation.Can you guide
the correct approach here?
3.) Choosing FuzzyLookupFactory looks beneficial, but we have to filter the
results over user context and also we need to provide infix search
capabilities for suggestions which we can't.
Can some one please help on this? Thanks in advance.
Harsha.
Re: Using n-grams vs AnalyzingInfixLookupFactory for suggestions in solr
Posted by Alessandro Benedetti <ab...@apache.org>.
Hi Harsha,
my blog can help :
http://alexbenedetti.blogspot.co.uk/2015/07/solr-you-complete-me.html
Take a look to the related section.
Cheers
On Sun, Jun 26, 2016 at 5:27 PM, Erick Erickson <er...@gmail.com>
wrote:
> 1> The difference is that the factory returns the original field
> from a "sidecar" index. If you're content with just firing a standard
> query at your main index and returning the associated fields
> then you can do this from the main index. You won't be able
> to do the sophisticated stuff with weights though.
>
> 2> https://issues.apache.org/jira/browse/LUCENE-6336
>
> 3> You can have multiple suggesters in the same component.
>
> Best,
> Erick
>
> On Thu, Jun 23, 2016 at 10:21 PM, Harsha JSN <ha...@gmail.com> wrote:
> > Hi,
> > I have some doubts regarding usage of AnalyzingInfixLookupFactory as
> > lookup implementation for suggestions.
> >
> > 1.) AnalyzingInfixLookupFactory constructs n-grams for the suggestion
> field
> > while building suggestions index. If the main index which is used for
> > search is already having n-grams for this field, is it still preferred to
> > choose AnalyzingInfixLookupFactory or can we directly build suggestions
> > from the main index?
> >
> > 2.) Also, AnalyzingInfixLookupFactory returns duplicate records if the
> > suggestion field has same value in multiple documents. Instead if i
> search
> > suggestions from main index (n-grams) i can eliminate the duplicates by
> > grouping the results. But grouping can be a complex operation.Can you
> guide
> > the correct approach here?
> >
> > 3.) Choosing FuzzyLookupFactory looks beneficial, but we have to filter
> the
> > results over user context and also we need to provide infix search
> > capabilities for suggestions which we can't.
> >
> > Can some one please help on this? Thanks in advance.
> >
> > Harsha.
>
--
--------------------------
Benedetti Alessandro
Visiting card : http://about.me/alessandro_benedetti
"Tyger, tyger burning bright
In the forests of the night,
What immortal hand or eye
Could frame thy fearful symmetry?"
William Blake - Songs of Experience -1794 England
Re: Using n-grams vs AnalyzingInfixLookupFactory for suggestions in solr
Posted by Erick Erickson <er...@gmail.com>.
1> The difference is that the factory returns the original field
from a "sidecar" index. If you're content with just firing a standard
query at your main index and returning the associated fields
then you can do this from the main index. You won't be able
to do the sophisticated stuff with weights though.
2> https://issues.apache.org/jira/browse/LUCENE-6336
3> You can have multiple suggesters in the same component.
Best,
Erick
On Thu, Jun 23, 2016 at 10:21 PM, Harsha JSN <ha...@gmail.com> wrote:
> Hi,
> I have some doubts regarding usage of AnalyzingInfixLookupFactory as
> lookup implementation for suggestions.
>
> 1.) AnalyzingInfixLookupFactory constructs n-grams for the suggestion field
> while building suggestions index. If the main index which is used for
> search is already having n-grams for this field, is it still preferred to
> choose AnalyzingInfixLookupFactory or can we directly build suggestions
> from the main index?
>
> 2.) Also, AnalyzingInfixLookupFactory returns duplicate records if the
> suggestion field has same value in multiple documents. Instead if i search
> suggestions from main index (n-grams) i can eliminate the duplicates by
> grouping the results. But grouping can be a complex operation.Can you guide
> the correct approach here?
>
> 3.) Choosing FuzzyLookupFactory looks beneficial, but we have to filter the
> results over user context and also we need to provide infix search
> capabilities for suggestions which we can't.
>
> Can some one please help on this? Thanks in advance.
>
> Harsha.