BIRD: Bronze Inscription Restoration and Dating

TL;DR

This paper presents BIRD, a new dataset and a specialized language model that improves the restoration and dating of fragmentary Bronze inscriptions from early China by integrating domain knowledge and glyph information.

Contribution

The paper introduces BIRD, a comprehensive dataset and a novel allograph-aware language model with Glyph Net for better restoration and dating of bronze inscriptions.

Findings

Glyph Net enhances restoration accuracy

Glyph-biased sampling improves dating performance

Allograph-aware modeling benefits inscription analysis

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.