Intelligent Character Recognition of Arabic Manuscripts in Africa: a Sample from the Manuscripts of Imam Muhammad ibn Abdul Karim Al-Maghili

Abstract

Abstract: This study experimented with the Zinki application for the intelligent recognition of Arabic letters in the manuscript of Imam Muhammad ibn Abdul Karim al-Maghili. The manuscript is a one-page letter containing 23 lines of unknown date, addressed to the people of Touat (currently the city of Adrar - Algeria). The aim of the study is to assess the ability and accuracy of the Zinki application in recognizing and extracting Arabic letters and words from Arabic manuscripts into editable and searchable text. The result showed an accuracy rate of 94.8% for letter recognition and 91.9% for word recognition. However, after analysis and comparison between the extracted text and the manuscript, we noticed several errors in the application, indicating that these percentages do not accurately reflect the application's capability to handle different types and forms of Arabic script and the distortions of Arabic letters. Keywords : Intelligent character recognition; Arabic script; Zinki application; Imam al-Maghili.