Abstract
Copy detection in Digital Libraries may provide the necessary guarantees for publishers and newsfeed ser¬vices to o~er valuable on-line data. We consider the case for a registration server that maintains regis¬tered documents against which new documents can be checked for overlap. In this paper we present a new scheme for detecting copies based on compar¬ing the word frequency occurrences of the new docu¬ment against those of registered documents. We also report on an experimental comparison between our proposed scheme and COPS [6], a detection scheme based on sentence overlap. The tests involve over a million comparisons of netnews articles and show that in general the new scheme performs better in detecting documents that have partial overlap.
