In this paper, we study the effects of automatic zoning on retrieval and ranking variability. We will show that OCR-generated text from automatic zoning, followed by postprocessing, produces retrieval results equivalent to OCR-generated text from manual zoning. We further show that there is a strong linear association between the ranked query results obtained from these two methods of zoning.
Received: 17 July 2003, Accepted: 18 October 2003, Published online: 6 February 2004Information Science Research Institute: e-mail isri@isri.unlv.edu