A Review of Artificial Intelligence Techniques in Oracle Bone Inscriptions

Oracle Bone Inscriptions (OBIs), incised over 3,000 years ago, represent one of humanity’s earliest extensively attested writing systems and preserve invaluable records of early Chinese civilization. Despite their status as UNESCO Memory of the World Register, OBI research still faces severe challenges, including extreme data scarcity, pervasive physical degradation, and persistent semantic ambiguity because many characters remain undeciphered. This paper reviews recent artificial intelligence (AI) techniques for OBI studies across three domains: organization and digitization for authentication and periodization, character-level detection and retrieval, and higher-level paleographic analysis, including semantic interpretation and known-class recognition support. We review representative approaches and publicly available datasets, highlight key technical bottlenecks such as data scarcity, weak cross-domain generalization, and semantic gaps between visual matching and cultural interpretation, and outline future directions, including robust models for degraded materials, methods that generalize across diverse collections, and approaches that integrate computational efficiency with paleographic expertise. By integrating computational methods with traditional paleography, this review shows how AI can accelerate OBI curation and enable large-scale cultural heritage analysis.