Yes, stemmed searches use language-specific keys in the index, so the IDFs of 
these are calculated independently.

//Mary

On 12/15/2016 08:34 AM, Andreas Hubmer wrote:
Hi,

It seems that the weight is - at least partly - language dependent.
After changing the xml:lang at the root element of a document, the weight of 
the term in the relevance info of that document changes as well.

I'm searching over English and German documents using an or-query:
cts:or-query((
  cts:field-word-query("myfield", $searchterm, "lang=en"),
  cts:field-word-query("myfield", $searchterm, "lang=de")
))

Is the term weight in the relevance info related to the 
inverse-document-frequency (IDF)?
Could it be that the IDF of a term is calculated separately in each language?

How can I prevent that this leads to a boost for documents in one of then 
languages?
Example: A user searches for "brain" which occurs in a lot of English documents 
but very little German documents. To me it seems that the German documents 
containing "brain" are ranked higher because of scoring with IDF.

Regards,
Andreas


2016-12-15 16:36 GMT+01:00 Andreas Hubmer 
<[email protected]<mailto:[email protected]>>:
Hi,

I'm tuning the result order of a search and not sure about all the parts in the 
output of cts:relevance-info.

Example:
<qry:relevance-info>
  ...
  <qry:term weight="47.75">
    <qry:score formula="8*weight*logtf" computation="382*20">7640</qry:score>
    <qry:key>12437021743613916800</qry:key>
  </qry:term>
</qry:relevance-info>

What is the weight marked in bold? How is it influenced?
The term query in my example is a field-word-query.

Thanks,
Andreas

--
Andreas Hubmer
Senior IT Consultant

EBCONT enterprise technologies GmbH
Millennium Tower
Handelskai 94-96
A-1200 Vienna




_______________________________________________
General mailing list
[email protected]<mailto:[email protected]>
Manage your subscription at:
http://developer.marklogic.com/mailman/listinfo/general


_______________________________________________
General mailing list
[email protected]
Manage your subscription at: 
http://developer.marklogic.com/mailman/listinfo/general

Reply via email to