PurposeThe authors investigate optical character recognition (OCR) technology and discuss its implementation in the context of digitisation of archival materials.Design/methodology/approachThe typewritten transcripts of the Croatian Writers' Society from the mid-60s of the 20th century are used as the test data. The optimal digitisation setup is investigated in order to obtain the best OCR results. This was done by using the sample of 123 pages digitised at different resolution settings and binarisation levels.FindingsA series of tests showed that different settings produce significantly different results. The best OCR accuracy achieved at the test sample of the typewritten documents was 95.02%. The results show that the resolution is significantly more important than binarisation pre-processing procedure for achieving better OCR results.Originality/valueBased on the research results, the authors give recommendations for achieving optimal digitisation process setup with the aim of increasing the quality of OCR results. Finally, the authors put the research results in the context of digitisation of cultural heritage in general and discuss further investigation possibilities.

The purpose of this research is to establish a method for OCR quality evaluation in different archival situations and stages of document ingest process. The conducted experiments explain the importance of OCR optimization using the example of Croatian typewritten materials. Furthermore, the paper discusses the importance of unique, or distinct, words in document retrieval.