An Optical Character Recognition Software Benchmark for Old Dutch Texts on the EYRA Platform

Digitized collections of printed historical texts are important for research in Digital Humanities.
However, acquiring high-quality machine readable texts using currently available Optical Character
Recognition (OCR) methods is a challenge. OCR Quality is affected by old fonts, old printing
techniques, bleedthrough of the ink, paper quality, old spelling, multiple columns and so on. It is
unclear which OCR methods perform best. Therefore, we are currently in the process of setting up a
benchmark to enable the evaluation of the performance of OCR software on old Dutch texts. The
benchmark is being set-up on the EYRA benchmark platform (eyrabenchmark.net) developed by The
Netherlands eScience Center and SURF.