Pegon AI: Toward A Smart Transliteration System For The Preservation Of Pegon Manuscripts

Abstrak

The Pegon script is a valuable cultural and religious literacy heritage within Islamic communities in Indonesia. However, its usage has significantly declined, primarily due to the dominance of the Latin alphabet, which has been extensively digitized. Various preservation efforts such as the development of transliteration tools and virtual keyboards have been introduced, yet most still face limitations in terms of linguistic accuracy, contextual understanding, and user accessibility. To date, no system has been available that can transliterate Pegon text from images either through takeing a photo or selecting from gallery and there is no integration within word processing software to facilitate efficient and accurate digital writing in Pegon script. This study proposes the development of Pegon AI, an intelligent transliteration system powered by artificial intelligence, consisting of two main components: (1) a mobile application that supports image based transliteration of Pegon script through photo capture and gallery selection, and (2) a Microsoft Word add ins capable of accurately converting Latin script to Pegon, thereby supporting culturally contextual digital writing practices. The system utilizes Computer Vision to identify the positions of text lines in images and employs a Vision Language Model (VLM) to perform accurate and context aware transliteration. This research adopts a qualitative methodology, employing case studies and participatory observation to understand user needs and evaluate the potential of the system within relevant social and cultural settings. Such an approach provides a strong foundation for the development of a transliteration system that is adaptive, user oriented, and capable of supporting the preservation of the Pegon script in the digital era.