Highlighter highlighting terms which are not part of the search

[Shyam Bhaskaran <Sh...@synopsys.com>]

Hi

We recently upgraded our Solr to the latest 4.0 trunk and we are seeing a weird behavior with highlighting which was not seen earlier.

When a search query for example "generate test pattern" is passed in the results et obtained the first few results shows the highlighting properly but in the later results we see terms which were not part of the search like "Question", "Answer", "used" etc. are being highlighted. We are using regular and termVectorHighlighter and never faced this kind of scenario, edismax is used in our configuration.

Can someone point to what is causing this problem and where I need to look into for fixing this?

-Shyam

RE: Highlighter highlighting terms which are not part of the search

[Shyam Bhaskaran <Sh...@synopsys.com>]

Hi Erick,

I tried looking into our analyzers and also adding each of the filters that we were using one by one and getting the documents indexed and during this testing it was found that when using the "solr.SynonymFilterFactory" on top of the latest Solr 4.0 trunk code there is issue with highlighting.
Some unwanted terms which are not part of the search are getting highlighted. This issue has come up after using the latest Solr4.0 trunk, earlier the search and highlighting was working fine. Looks like some issue with SynonymFilterFactory.

-Shyam


-----Original Message-----
From: Erick Erickson [mailto:erickerickson@gmail.com] 
Sent: Tuesday, December 13, 2011 7:51 PM
To: solr-user@lucene.apache.org
Subject: Re: Highlighter highlighting terms which are not part of the search

Well, we need some more details to even guess.
Please review:
http://wiki.apache.org/solr/UsingMailingLists

Best
Erick


On Mon, Dec 12, 2011 at 12:04 AM, Shyam Bhaskaran
<Sh...@synopsys.com> wrote:
> Hi
>
> We recently upgraded our Solr to the latest 4.0 trunk and we are seeing a weird behavior with highlighting which was not seen earlier.
>
> When a search query for example "generate test pattern" is passed in the results et obtained the first few results shows the highlighting properly but in the later results we see terms which were not part of the search like "Question", "Answer", "used" etc. are being highlighted. We are using regular and termVectorHighlighter and never faced this kind of scenario, edismax is used in our configuration.
>
> Can someone point to what is causing this problem and where I need to look into for fixing this?
>
> -Shyam

Re: Highlighter highlighting terms which are not part of the search

[Erick Erickson <er...@gmail.com>]

Well, we need some more details to even guess.
Please review:
http://wiki.apache.org/solr/UsingMailingLists

Best
Erick


On Mon, Dec 12, 2011 at 12:04 AM, Shyam Bhaskaran
<Sh...@synopsys.com> wrote:
> Hi
>
> We recently upgraded our Solr to the latest 4.0 trunk and we are seeing a weird behavior with highlighting which was not seen earlier.
>
> When a search query for example "generate test pattern" is passed in the results et obtained the first few results shows the highlighting properly but in the later results we see terms which were not part of the search like "Question", "Answer", "used" etc. are being highlighted. We are using regular and termVectorHighlighter and never faced this kind of scenario, edismax is used in our configuration.
>
> Can someone point to what is causing this problem and where I need to look into for fixing this?
>
> -Shyam