[Topic-models] What about introducing a weight for words in a document?

Marie Lienou marie.lienou at telecom-paristech.fr
Wed Nov 19 05:19:43 EST 2008


Hi all,
My question is about introducing a weight for words in a document. For
example, I'm working with images, and a word represents a document. I
consider that a word is a region of the image. But all the regions of
the image don't have the same size for example, so just counting the
occurrences of the different words might not be right. I saw that in
some works, people also use geometrical features of the regions before
performing the quantization when constructing the vocabulary. But is it
possible to introduce a weight for each word in the bags-of-words
representation instead of just counting the occurrences of the different
words? Do you have any idea about papers related to this problem?
Thanks,
Marie

-- 
Marie LIENOU
Ph.D. student
TELECOM ParisTech, TSI Department
46 rue Barrault - 75 013 Paris - FRANCE 
Phone : +33 (0)1 45 81 73 91
e-mail : marie.lienou at telecom-paristech.fr



More information about the Topic-models mailing list