Conversion of card catalogues to a computerized catalogue, called retrospective conversionrnor RECON, is the primary hurdle of effective library automation. There are several optionsrnfor RECON, but recent advances in OCR technology make it a prominent alternative. Thisrnthesis has mainly examined technical feasibility of the OCR technology for RECON withrnparticular reference to Addis Ababa University libraries' union catalogue The thesis reviews the various options for RECON in the context of Addis Ababa UniversityrnLibrary System. It also reviews technological advancements of optical scanning and OCRrntechnology stressing on document scanning and OCR with particular reference to theirrnapplication in libraries, particularly RECON. An introduction to the Addis Ababa UniversityrnLibrary System gives an overview of the library system with particular reference to itsrncataloguing department and cataloguing practices followed therein, and automation plan ofrnthe library system.Two sets of catalogue cards, each consisting 115 added up to 230, were chosen from thernunion catalogue for this study. The sets of cards consisted of main entries made under namernof personal author (40 each), under title (40 each), and corporate body (35 each). Arnprototype program, called P AEBE, was written in C++ using the first set of cards. Thernvarious steps include: scanning and conversion of the cards to ASCII text files, analyzing ofrnthe files, manual editing of the files, called reprocessing, if necessary, writing and runningrnthe program creating output records consisting of each bibliographic item preceded by an appropriate field identifier.rnIt was noted that the success of the prototype depends much on the accuracy of OCR software,rnconsistency of information on card catalogues, and quality of card catalogues. The prototypernwas tested with the second set of catalogue cards; and the results are discussed. It was foundrnout that the performance of the prototype program is encouraging. The thesis also highlightsrnimplementation strategy for the RECON using OCR technology. Finally, some conclusions andrnrecommendations for further studies are forwarded.