OCR quality: Key to enhanced Data Mining

Quality of scanned documents vary, that is a fact. Another fact is that without high quality optical character recognition (OCR) it is difficult or impossible to achieve good usability or findability. This paper focuses on improving the OCR quality by performing different pre-OCR image manipulations and testing the performance of different OCR engines. The results clearly indicate that there are differences between the performance of the OCR engines, but the pre-OCR image manipulations did not significantly affect the results. Tests were conducted with mostly restricted content which came from the National Archives of Finland. The figures shown in this paper are based on tests where material from the Central Archives of Finnish Business Records was used.