help-gift
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[help-GIFT] Pruning continued


From: Wolfgang Mueller
Subject: [help-GIFT] Pruning continued
Date: Mon, 3 Sep 2001 23:17:20 +0200

I wrote some while ago:
> It means
> the program will read the feature list for the 70% query features with the
> highest term frequency.


Oops, this is wrong. Thanks, Henning, for pointing out there was a bug in my 
message. Here's the fix: the truth can be found in this line,  taken from 
CQInvertedFile.cc, line 650, to be precise:

//[ QFW stands for QueryFeatureWeighters ]

outQFW.sort(CSortPointers_WF <CSortByAbsQueryFactor_WF>());   

In fact, we sort things by the absolute value of the "query factor" before 
pruning. The query factor contains all the knowledge about the weight that 
depends only on the query (and not on the retrieved document, e.g. the term 
frequency of the feature in the retrieved document).  In the case of a tf.idf 
weight this will be tf(query).idf with tf(query) the effective term 
frequency. I hope this time I got it right, please don't hesitate to correct 
me.

BTW: See CWeightingFunction.h for more descriptions about the query factor.

Cheers,
Wolfgang
-- 
Wolfgang M&uuml;ller, 
assistant-doctorant ==  PhD student (2001), teaching assistant
Personal page: http://cui.unige.ch/~vision/members/WolfgangMueller.html 
Maintainer, GNU Image Finding Tool (http://www.gnu.org/software/gift)




reply via email to

[Prev in Thread] Current Thread [Next in Thread]