lucene-solr-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Lasitha Wattaladeniya <watt...@gmail.com>
Subject Re: Implementing partial search and exact matching
Date Fri, 06 May 2016 03:25:05 GMT
Hi Nick,

Thanks for the reply.  Actually

q="software engineering" - > doc1
q="software engineer" - > no results
q="Software engineer" - > doc2

I hope above test cases will explain my requirements further.  So far I'm
thinking of changing the qf according to query has enclosing double
quotations or not.

If somebody knows a better approach/queryparser please point me

Thanks
On 6 May 2016 10:56 am, "ND" <ndrake0027@gmail.com> wrote:

> Lasitha,
>
> I think I understand what you are asking and if you have something like
> Doc1 = software engineering
> Doc2 = Software engineer
>
> And if you query
> q=software+engineer -> Doc1 & Doc2
>
> but
>
> q="software+engineer" -> Doc1
>
> Correct?
>
> If this is correct then to my knowledge no, Solr out of the box cannot do
> what you are asking, that is recognize a exact (quoted) search and change
> the query fields to non-ngram fields. Of course you could do this in code
> with some regex to see if the first and last character are double quotes
> but there are a number of draw backs to this like q= "Software Engineer"
> and Ninja wont do what you want it to but you could always do something
> else.
>
> From my experience I would still want q="software+engineer" -> Doc1 and
> Doc2 cause technically that exact phrase does exists in both Docs.
>
> Maybe there is someone else on here who can offer some more perspective on
> this or a possible query analyzer that I haven't heard of that can solve
> this issue (I would also be interested in that).
>
> Nick
>
> On Thu, May 5, 2016 at 6:33 PM, Lasitha Wattaladeniya <wattale@gmail.com>
> wrote:
>
> > Hi nd,
> >
> > Here's the issue..  Let's say I search.. Software Engineer. For am
> example
> > lets say this query will return 10 results when search against ngram
> field
> > . Now I search "Software Engineer"   with double quotations. This should
> > not return same result set as the previous query.
> >
> > I thought the query parser I'm using (edismax)  may have an inbuilt
> > function for that.
> >
> > Do I have to specifically change the query field (qf)  to solve this
> issue
> > for each query?
> > On 6 May 2016 8:26 am, "ND" <ndrake0027@gmail.com> wrote:
> >
> > We implemented something similar it sounds to what you are asking but I
> > dont see why you would need to match the original field. Since
> technically
> > a field that has *software engineer* indexed is matched by a query like
> > "software eng" .... to "software engineer" with the ngrams; which makes
> the
> > exact phrase is still valid.
> >
> > The problem we where trying to solve was the exact phrase issue which can
> > be solved with taking in a qs value of 0 (or higher depending on your
> > definition of exactness).
> >
> > Maybe an example would help if there is something I am not understanding.
> >
> > Also your field definitions might help or a example of the schema
> breakdown
> > similar to the admin analyze page.
> >
> > Nick
> >
> >
> > On Thu, May 5, 2016 at 12:25 AM, Lasitha Wattaladeniya <
> wattale@gmail.com>
> > wrote:
> >
> > > Hi All,
> > >
> > > I'm trying to implement a search functionality using solr. Currently
> I'm
> > > suing edismax parser with ngram fields to do the search against. So far
> > it
> > > works well.
> > >
> > > The question I have is when the user input double quotations to the
> > search,
> > > As the requirement this should  match against the original field, not
> > > against the ngram field.
> > >
> > > Currently what I have thought of doing is, identify the double
> quotations
> > > in the user input and change the query field (qf) according to that (to
> > > ngram field or to the exact field). Isn't there any out of the box
> > solution
> > > for this, I feel like it's a generic requirement and don't want to
> > reinvent
> > > the wheel. Appreciate your comments
> > >
> > > [1]. https://issues.apache.org/jira/browse/SOLR-6842
> > >
> > > [2].
> > >
> > >
> >
> >
> http://grokbase.com/t/lucene/solr-user/14cbghncvh/different-fields-for-user-supplied-phrases-in-edismax
> > >
> > > Thanks,
> > >
> > > Lasitha Wattaladeniya
> > >
> > > Software Engineer
> > >
> >
>

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