Making Arabic PDF books accessible using gamification

Abstract

Most of online Arabic books are not accessible to Arab people with visual impairments. They cannot read online books because they are usually scanned images of the original ones. There is also a problem in PDF encoding of some of the textual books. One of the solutions is to use an Arabic OCR to convert scanned books into text; however Arabic OCR is still in its early stages and suffers from many limitations. In this paper we propose the use of human recognition skills to replace OCR limitations by incorporating the concepts of crowdsourcing and gamification. Our proposed system is in the form of a mobile recall game that presents players with word images segmented from the books to be converted into text. The players' answers are checked using techniques similar to what is used in word spotting. We initially implemented two components of the system; the segmentation, and the feature extraction and matching components. For the feature extraction and matching component, which is used to verify the player's answers, we performed four tests to choose a similarity measure threshold for accepting an entered word as a correct answer. Future work will consider other means of input correctness assurance.