BIRD: Bronze Inscription Restoration and Dating

2025-11-03

AI Summary

To address the challenges of fragmented inscriptions and uncertain dating in early Chinese bronze inscriptions, this paper constructs the first fully encoded dataset featuring standardized transcription and precise chronological annotation, and proposes a glyph-aware joint restoration and dating framework. Methodologically, it introduces (1) a Glyph Net that models the diachronic evolution of bronze script glyphs via radical-variant mapping, and (2) variant-aware encoding coupled with glyph deviation sampling to enhance domain-adaptive pretraining. Experimental results demonstrate that Glyph Net significantly improves inscription restoration accuracy, while glyph deviation sampling markedly enhances dating performance. The joint framework achieves substantial gains on both tasks. This work establishes a scalable methodological foundation for intelligent processing of ancient Chinese characters, advancing computational paleography through integrated structural and temporal modeling of script evolution.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Search and Optimization: Evolutionary Computation
• Machine Learning: Neuro-Symbolic Learning

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Web Mining and Content Analysis: Models for Web evolution
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Bronze inscriptions from early China are fragmentary and difficult to date. We introduce BIRD(Bronze Inscription Restoration and Dating), a fully encoded dataset grounded in standard scholarly transcriptions and chronological labels. We further propose an allograph-aware masked language modeling framework that integrates domain- and task-adaptive pretraining with a Glyph Net (GN), which links graphemes and allographs. Experiments show that GN improves restoration, while glyph-biased sampling yields gains in dating.

Problem

• Restores fragmented bronze inscriptions from early China
• Determines accurate dating of ancient bronze artifacts
• Links graphemes and allographs through Glyph Net framework

Innovation

• Uses allograph-aware masked language modeling framework
• Integrates domain-adaptive pretraining with Glyph Net
• Links graphemes and allographs for improved restoration