Are query scores ordinal for different documents across queries?


(Ilija Subasic) #1

I am trying to figure out whether query scores are ordinal or not. Trying
to figure this out, but have a problem with the effects coord function in
Lucene's forumla would do. Basically, if I have query A that says document
X is relevant with some score W, and query B which is relevant to document
Y with some score V, and W > V can I say that document X is more related A,
than a document Y is related to B.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/775c8d49-fb62-4132-acf1-08ebe8b208ad%40googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.


(Ivan Brusic) #2

Query scores are not ordinal. Lucene discourages comparing scores from
different queries since the search context is important in scoring. Query
norms are supposed to help normalize scores across queries, but as you
discovered, that is not normally what happens. Your theory will probably
hold true for many cases, but it is not 100% correct in all cases.

--
Ivan

On Thu, Jan 30, 2014 at 7:06 AM, Ilija Subasic subasic.ilija@gmail.comwrote:

I am trying to figure out whether query scores are ordinal or not. Trying
to figure this out, but have a problem with the effects coord function in
Lucene's forumla would do. Basically, if I have query A that says document
X is relevant with some score W, and query B which is relevant to document
Y with some score V, and W > V can I say that document X is more related A,
than a document Y is related to B.

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/elasticsearch/775c8d49-fb62-4132-acf1-08ebe8b208ad%40googlegroups.com
.
For more options, visit https://groups.google.com/groups/opt_out.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/CALY%3DcQA6w78ZxqDc%2BDvt7%2Bq9mKQn-KiLTpGz6-8awQtppgSm0w%40mail.gmail.com.
For more options, visit https://groups.google.com/groups/opt_out.


(Ilija Subasic) #3

Thanks for confirming my suggestions. Even though the statement would stand
in general, since it is not ordinal it can only produce ranking, right?
So, if I would change the scoring function to be simple cosine of tf.idf
weights it should be ordinal. Can I do this in MLT query?
Thanks again

On Friday, January 31, 2014 12:42:38 AM UTC+1, Ivan Brusic wrote:

Query scores are not ordinal. Lucene discourages comparing scores from
different queries since the search context is important in scoring. Query
norms are supposed to help normalize scores across queries, but as you
discovered, that is not normally what happens. Your theory will probably
hold true for many cases, but it is not 100% correct in all cases.

--
Ivan

On Thu, Jan 30, 2014 at 7:06 AM, Ilija Subasic <subasi...@gmail.com<javascript:>

wrote:

I am trying to figure out whether query scores are ordinal or not. Trying
to figure this out, but have a problem with the effects coord function in
Lucene's forumla would do. Basically, if I have query A that says document
X is relevant with some score W, and query B which is relevant to document
Y with some score V, and W > V can I say that document X is more related A,
than a document Y is related to B.

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearc...@googlegroups.com <javascript:>.
To view this discussion on the web visit
https://groups.google.com/d/msgid/elasticsearch/775c8d49-fb62-4132-acf1-08ebe8b208ad%40googlegroups.com
.
For more options, visit https://groups.google.com/groups/opt_out.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/6a1d27e7-fcc3-47c4-9447-330555235bcb%40googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.


(system) #4