Two field phrase search
Hi Fellow Searchers,
I am indexing some content (say name) in to two fields with different
analyzer chains. One that would do edge prefix and the other regular
tokenization.
Let me call them name_prefix and name_term
When an incoming partial query comes up, like "steve jo", I want the
first word to be searched in regular index while the jo to go to
prefix index.
something like
name_term:steve AND name_prefix:jo
works.
But I would like to use them in a phrase. i.e. expect them to be next
to each other. Following are the things attempted. with out success.
1. regular phrase: "name_term:steve name_prefix:jo" - does not expect
field names.
2. complex phrase query: {!complexphrase inOrder=true}"name_term:steve
name_prefix:jo" - does not allow different field names in the same
phrase.
3. played around with {!surround}3w(name_term:steve,name_prefix:jo) -
does not support subquery (field names).
Any other way this can be accomplished?
Thanks
Melchi
Re: Two field phrase search
Thanks Erick for your inputs,
Yes, i was indeed looking for this. I was little worried about the *
expansion. As it is already part of a large query. But this sounds like
the simplest way I can achieve this for now.
If I indeed slows down, will revisit this.
Was wondering if some how I can create a spanterm from different fields, I
should be able to position limit them using span aware parsers. Sorry, if
I am just day dreaming.
Thanks
Melchi
On Fri, 23 Nov 2018 at 08:18, Erick Erickson
wrote:
> There's no capability in Solr to do this. Given that different fields
> have different analyzer chains, how would position information be
> preserved across two fields? It'd be tricky at best.
>
> ComplexPhrase should do what you want by itself, assuming you want
> matches for any doc with "steve jobs", "steve johns", "steve johnson"
> etc.. Just search for name_term:"steve jo*"
>
> Best,
> Erick
> On Thu, Nov 22, 2018 at 4:37 PM NewMelchizedec sundararaj
> wrote:
> >
> > Hi Fellow Searchers,
> >
> > I am indexing some content (say name) in to two fields with different
> > analyzer chains. One that would do edge prefix and the other regular
> > tokenization.
> > Let me call them name_prefix and name_term
> >
> > When an incoming partial query comes up, like "steve jo", I want the
> > first word to be searched in regular index while the jo to go to
> > prefix index.
> >
> > something like
> > name_term:steve AND name_prefix:jo
> >
> > works.
> >
> > But I would like to use them in a phrase. i.e. expect them to be next
> > to each other. Following are the things attempted. with out success.
> >
> > 1. regular phrase: "name_term:steve name_prefix:jo" - does not expect
> > field names.
> > 2. complex phrase query: {!complexphrase inOrder=true}"name_term:steve
> > name_prefix:jo" - does not allow different field names in the same
> > phrase.
> > 3. played around with {!surround}3w(name_term:steve,name_prefix:jo) -
> > does not support subquery (field names).
> >
> > Any other way this can be accomplished?
> >
> > Thanks
> > Melchi
>
--
Re: Two field phrase search
Hi Alex,
Thanks for your reply. I am keeping the original with the edge, but not
sure how to mark them as original. I would like to avoid a prefix match
for the full words.
You mentioned post-filter - can you shed some light on this?. If only I
can remove these over-matches, that would be great.
My requirement is even more involved than my question. say I have a set of
analyser chains
1. token
2. edgeNgram
3. phonetic
etc
I wanted to match every token against say these analyzed fields with
different weight, but never double counting their match for the same
token. So I am wrapping them in a {!maxscore}.
In addition to this, I would like to maintain the proximity of the tokens
to one another (at least maintaining the proximity less than the
positionIncrementGap to remain with in a single value of a multi-valued
field).
Regards
Melchi
On Fri, 23 Nov 2018 at 17:54, Alexandre Rafalovitch
wrote:
> It is not clear how much flexibility you expect in those queries. Can
> the second word never be full name? Can there be more than 2 words?
> How do you know the length of the prefix? When you say prefix, do you
> mean 'jo' is expected to match 'joseph'?
>
> So, just generically, I would say why not index everything into one
> field with analyzer chain keeping both original and prefixed
> (EdgeNGrams) forms of the word? Maybe combined with some sort of
> post-filter if that over-matches.
>
> Regards,
> Alex.
> On Thu, 22 Nov 2018 at 19:37, NewMelchizedec sundararaj
> wrote:
> >
> > Hi Fellow Searchers,
> >
> > I am indexing some content (say name) in to two fields with different
> > analyzer chains. One that would do edge prefix and the other regular
> > tokenization.
> > Let me call them name_prefix and name_term
> >
> > When an incoming partial query comes up, like "steve jo", I want the
> > first word to be searched in regular index while the jo to go to
> > prefix index.
> >
> > something like
> > name_term:steve AND name_prefix:jo
> >
> > works.
> >
> > But I would like to use them in a phrase. i.e. expect them to be next
> > to each other. Following are the things attempted. with out success.
> >
> > 1. regular phrase: "name_term:steve name_prefix:jo" - does not expect
> > field names.
> > 2. complex phrase query: {!complexphrase inOrder=true}"name_term:steve
> > name_prefix:jo" - does not allow different field names in the same
> > phrase.
> > 3. played around with {!surround}3w(name_term:steve,name_prefix:jo) -
> > does not support subquery (field names).
> >
> > Any other way this can be accomplished?
> >
> > Thanks
> > Melchi
>
--
Re: Two field phrase search
Thanks Mikhail,
I will check this FieldMaskingSpanQuery out.
Thanks
Melchi
On Fri, 23 Nov 2018 at 19:17, Mikhail Khludnev wrote:
>
> There's no way to search with positions across other fields. I can only
> recommend to prepend prefix terms to distinguish them.
> But hold on, what about FieldMaskingSpanQuery? It's available in
> XMLQparser, you may try.
>
> On Fri, Nov 23, 2018 at 3:37 AM NewMelchizedec sundararaj
> wrote:
>
> > Hi Fellow Searchers,
> >
> > I am indexing some content (say name) in to two fields with different
> > analyzer chains. One that would do edge prefix and the other regular
> > tokenization.
> > Let me call them name_prefix and name_term
> >
> > When an incoming partial query comes up, like "steve jo", I want the
> > first word to be searched in regular index while the jo to go to
> > prefix index.
> >
> > something like
> > name_term:steve AND name_prefix:jo
> >
> > works.
> >
> > But I would like to use them in a phrase. i.e. expect them to be next
> > to each other. Following are the things attempted. with out success.
> >
> > 1. regular phrase: "name_term:steve name_prefix:jo" - does not expect
> > field names.
> > 2. complex phrase query: {!complexphrase inOrder=true}"name_term:steve
> > name_prefix:jo" - does not allow different field names in the same
> > phrase.
> > 3. played around with {!surround}3w(name_term:steve,name_prefix:jo) -
> > does not support subquery (field names).
> >
> > Any other way this can be accomplished?
> >
> > Thanks
> > Melchi
> >
>
>
> --
> Sincerely yours
> Mikhail Khludnev
--
