You are viewing a plain text version of this content. The canonical link for it is here.
Posted to java-user@lucene.apache.org by Itamar Syn-Hershko <it...@code972.com> on 2011/06/16 23:01:27 UTC
Coloring search results based on score?
Hi all,
Interesting question: is it possible to color search results in a
web-page based on their score? e.g. most relevant results in green, and
then different shades through orange, yellow, red and then white.
Theoretically, one could take the highest score and color based on
proximity / distribution, but the highest score can be invalid in itself.
I could bring in a hardcoded cutoff point, where results are not
considered relevant and not color any results if the highest score
doesn't go above that threshold (e.g. score = 1.0), but then again -
complex queries can yield relevant results with low scores.
So my question is: has anyone ever tackled this issue, and is this even
doable?
Thanks in advance!
Itamar.
---------------------------------------------------------------------
To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-user-help@lucene.apache.org
Re: Coloring search results based on score?
Posted by Itamar Syn-Hershko <it...@code972.com>.
Thanks. That's very abstract and old, but perhaps I could work something
out using this.
Any other pointers / opinions welcome...
Itamar.
On 17/06/2011 03:26, Andrzej Bialecki wrote:
> On 6/17/11 12:29 AM, Itamar Syn-Hershko wrote:
>> No, that was not what I meant.
>>
>>
>> I'm not interested in coloring the actual text, but in giving the user
>> an indication of how relevant the results are. Instead of displaying the
>> result score, I want to give some visual meter to show that. The highest
>> ranking result will be green if the it is really relevant to the query,
>> yellow if less and so on.
>>
>>
>> My question is if there's a good enough way to measure this - for
>> example if the first results is 20 times the score of the second, or
>> that's something I can't really do...
>
> What you describe reminds me of a special case of a more general
> visualization technique called Tilebars. Tilebars are an N x M
> histogram, where N rows correspond to N query terms (or clauses), and
> M columns represent buckets of score contributions distributed over
> the sections of the document. Each cell is then colored according to
> the portion of score that it contains.
>
> The coloring schema may be non-linear, and the document division into
> chunks may not be linear either.
>
> You can then collapse Tilebars if you wish in one or both dimensions,
> e.g.:
>
> * N x 1 represents relative score contributions from each query term
> * 1 x M represents the distribution of score contributions along the
> length of document
> * 1 x 1 represents the relative score of all clauses for the whole
> document, relative to other documents (the difference is in coloring)
>
> See this link for more details:
> http://people.ischool.berkeley.edu/~hearst/research/tilebars.html
>
---------------------------------------------------------------------
To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-user-help@lucene.apache.org
Re: Coloring search results based on score?
Posted by Andrzej Bialecki <ab...@getopt.org>.
On 6/17/11 12:29 AM, Itamar Syn-Hershko wrote:
> No, that was not what I meant.
>
>
> I'm not interested in coloring the actual text, but in giving the user
> an indication of how relevant the results are. Instead of displaying the
> result score, I want to give some visual meter to show that. The highest
> ranking result will be green if the it is really relevant to the query,
> yellow if less and so on.
>
>
> My question is if there's a good enough way to measure this - for
> example if the first results is 20 times the score of the second, or
> that's something I can't really do...
What you describe reminds me of a special case of a more general
visualization technique called Tilebars. Tilebars are an N x M
histogram, where N rows correspond to N query terms (or clauses), and M
columns represent buckets of score contributions distributed over the
sections of the document. Each cell is then colored according to the
portion of score that it contains.
The coloring schema may be non-linear, and the document division into
chunks may not be linear either.
You can then collapse Tilebars if you wish in one or both dimensions, e.g.:
* N x 1 represents relative score contributions from each query term
* 1 x M represents the distribution of score contributions along the
length of document
* 1 x 1 represents the relative score of all clauses for the whole
document, relative to other documents (the difference is in coloring)
See this link for more details:
http://people.ischool.berkeley.edu/~hearst/research/tilebars.html
--
Best regards,
Andrzej Bialecki <><
___. ___ ___ ___ _ _ __________________________________
[__ || __|__/|__||\/| Information Retrieval, Semantic Web
___|||__|| \| || | Embedded Unix, System Integration
http://www.sigram.com Contact: info at sigram dot com
---------------------------------------------------------------------
To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-user-help@lucene.apache.org
Re: Coloring search results based on score?
Posted by Itamar Syn-Hershko <it...@code972.com>.
No, that was not what I meant.
I'm not interested in coloring the actual text, but in giving the user
an indication of how relevant the results are. Instead of displaying the
result score, I want to give some visual meter to show that. The highest
ranking result will be green if the it is really relevant to the query,
yellow if less and so on.
My question is if there's a good enough way to measure this - for
example if the first results is 20 times the score of the second, or
that's something I can't really do...
On 17/06/2011 01:21, Mark Harwood wrote:
> See Highlighter's GradientFormatter
>
> Cheers
> Mark
>
>
> On 16 Jun 2011, at 22:01, Itamar Syn-Hershko wrote:
>
>> Hi all,
>>
>>
>> Interesting question: is it possible to color search results in a web-page based on their score? e.g. most relevant results in green, and then different shades through orange, yellow, red and then white.
>>
>>
>> Theoretically, one could take the highest score and color based on proximity / distribution, but the highest score can be invalid in itself.
>>
>>
>> I could bring in a hardcoded cutoff point, where results are not considered relevant and not color any results if the highest score doesn't go above that threshold (e.g. score = 1.0), but then again - complex queries can yield relevant results with low scores.
>>
>>
>> So my question is: has anyone ever tackled this issue, and is this even doable?
>>
>>
>> Thanks in advance!
>>
>>
>> Itamar.
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
>> For additional commands, e-mail: java-user-help@lucene.apache.org
>>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
> For additional commands, e-mail: java-user-help@lucene.apache.org
>
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-user-help@lucene.apache.org
Re: Coloring search results based on score?
Posted by Mark Harwood <ma...@yahoo.co.uk>.
See Highlighter's GradientFormatter
Cheers
Mark
On 16 Jun 2011, at 22:01, Itamar Syn-Hershko wrote:
> Hi all,
>
>
> Interesting question: is it possible to color search results in a web-page based on their score? e.g. most relevant results in green, and then different shades through orange, yellow, red and then white.
>
>
> Theoretically, one could take the highest score and color based on proximity / distribution, but the highest score can be invalid in itself.
>
>
> I could bring in a hardcoded cutoff point, where results are not considered relevant and not color any results if the highest score doesn't go above that threshold (e.g. score = 1.0), but then again - complex queries can yield relevant results with low scores.
>
>
> So my question is: has anyone ever tackled this issue, and is this even doable?
>
>
> Thanks in advance!
>
>
> Itamar.
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
> For additional commands, e-mail: java-user-help@lucene.apache.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-user-help@lucene.apache.org