<![CDATA[This study examines the Sundanese Arab Pegon manuscript Wawacan Abdul Muluk through a philological approach to formulate a conceptual visual feature taxonomy framework as a foundational step for developing Artificial Intelligence (AI)-based Handwritten Text Recognition (HTR). The philological approach is employed to identify the manuscript’s condition, structural content, linguistic characteristics, and script, which are then transformed into digital parameters. The study successfully formulates six categories of conceptual visual features: paleography, text layout, script form, grapheme variation, ornamentation, and physical degradation (such as faded ink and damaged binding) that create visual noise affecting text legibility. These characteristics are analyzed as a baseline to enable AI models to accurately recognize, process, and represent Sundanese manuscripts. This synergy between philology and computational approaches contributes to the digital humanities by providing a visual feature taxonomy for future HTR application development. === Penelitian ini mengkaji manuskrip Arab Pegon Sunda Wawacan Abdul Muluk melalui pendekatan filologi untuk merumuskan kerangka kerja (framework) taksonomi fitur visual konseptual sebagai dasar awal pengembangan model Handwritten Text Recognition (HTR) berbasis Artificial Intelligence (AI). Pendekatan filologi digunakan untuk mengidentifikasi kondisi naskah, struktur isi, karakteristik kebahasaan, dan aksara, yang kemudian ditransformasikan menjadi parameter digital. Hasil penelitian berhasil merumuskan enam kategori fitur visual konseptual: paleografi, tata letak teks, bentuk aksara, variasi grafem, ornamen, serta degradasi fisik naskah (seperti tinta pudar dan jilid rusak) yang menjadi noise visual bagi keterbacaan teks. Karakteristik ini dianalisis sebagai landasan agar model AI mampu mengenali, memproses, dan merepresentasikan manuskrip Sunda secara akurat. Sinergi antara filologi dan pendekatan komputasional ini berkontribusi pada bidang humaniora digital dalam menyediakan taksonomi fitur visual untuk pengembangan aplikasi HTR di masa depan.]]>
