ISR10
Scientific Report No. ISR-10 Information Storage and Retrieval
Evaluation of Document Retrieval Systems
chapter
Joseph John Rocchio
Harvard University
Gerard Salton
Use, reproduction, or publication, in whole or in part, is permitted for any purpose of the United States Government.
5-22
define a function r*(x) equal to r*(i) for x =.i, a possible anomaly is
noted. This arises from the fact that it is possible, within the
framework of the system, for N to produce a mapping from elements of
q x to the real line. This, in fact, occurs when N is a correla-
tion process which correlates a query image with the set of document
images viewed as vectors in some abstract space. The process of
inducing an ordering from this mapping and then treating this ordering
as a function of a continuous real variable gives the impression of
coming full circle. In fact, there is clearly a loss of information
involved since relative distance between the images of d. and d. is not
1
preserved by this process. The justification for msking this transforma-
tion from the domain of N to an ordering index lies in the assumption
that the order so derived has significance of and by itself.
The extension then to functions of a real variable is
accomplished by defining two functions r*(x) and p*(x) such that:
r*(x) = r*(i)
> for x = i, i =
p*(x) = p*(i))
and further that:
( -[OCRerr] for j[OCRerr] x< j + 1 , and j integral and
less than n0
r*(x) =
1 for x >n
0
and I
- for j< x< j + 1., and j integral and
p*(x) =I[OCRerr]xn less than n0
0
-forx>n
0