COBHUNI at DATeCH Göttingen 2017
1 June 2017
COBHUNI will participate at the DATeCH conference, June 1st and 2nd 2017 in Göttigen with the paper
"Clear-cut methodology for Arabic OCR and post-correction with low technical skilled annotators"
This paper describes an efficient and straightforward methodology for OCR-ing and post-correcting Arabic text material on Islamic embryology collected for the COBHUNI project. As the target texts of the project include diverse diachronic stages of the Arabic language, the team of annotators for performing the OCR post-correction requires well-trained experts on language skills. While technical skills are also desirable, highly trained language experts typically lack enough technical knowledge. Furthermore, a relatively small portion of the target texts needed to be OCR-ed, as most of the material was already on some digital form. Thus, the OCR task could only require a small amount of resources in terms of time and work complexity. Both the low technical skills of the annotators and the resource constraints made it necessary for us to find an easy-to-develop and suitable workflow for performing the OCR and post-correction tasks. For the OCR phase, we chose Tesseract Open Source OCR Engine, because it achieves state-of-the-art levels of accuracy. For the post-correction phase, we decided to use the Proofread Page extension of the MediaWiki software, as it strikes a perfect balance between usability and efficiency. The post-correction task was additionally supported by the implementation of an error checker based on simple heuristics. The application of this methodology resulted in the successful and fast OCR-ing and post-correction of a corpus of 36,132 tokens.