Codicology of Islamic Manuscripts in Sumatra: The Impact of Deep Learning on Literature and Written Traditions

Abstract

This study examines the codicological richness of Islamic manuscripts in Sumatra and explores the transformative potential of deep learning technologies in their preservation. It aims to investigate how traditional written heritage, shaped by long-standing spiritual and intellectual legacies, can be revitalized through an interdisciplinary framework that integrates codicology, digital philology, and artificial intelligence. Employing a qualitative descriptive approach, the study analyzes twenty manuscripts from the DREAMSEA catalog, focusing on their physical characteristics, cultural contexts, and the role of digital technologies in enhancing accessibility and analytical depth. The findings reveal a diverse manuscript tradition across regions such as Ogan Komering Ilir, Padang Pariaman, and Sijunjung, encompassing texts on jurisprudence, Sufism, Qur’anic exegesis, and local customary practices. Codicological features including paper types, ink composition, scripts, and page layout provide critical insights for automated recognition and digital preservation processes. Deep learning models, particularly convolutional neural networks, demonstrate strong potential in supporting handwriting recognition, content classification, and visual restoration. Nevertheless, challenges persist in addressing script variability, faded ink, and limited digital literacy. The study concludes that while deep learning significantly enhances the efficiency of manuscript preservation, sustained human expertise and community engagement remain indispensable. A balanced approach that harmonizes technological innovation with cultural sensitivity is essential for safeguarding Islamic literary heritage in the digital age.