Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enabling Stroke-Level Structural Analysis of Hieroglyphic Scripts without Language-Specific Priors

Fuwen Luo, Zihao Wan, Ziyue Wang, Yaluo Liu, Pau Tong Lin Xu, Xuanjia Qiao, Xiaolong Wang, Peng Li, Yang Liu

TL;DR

Hieroglyphic Stroke Analyzer (HieroSA) is proposed, a novel and generalizable framework that enables MLLMs to automatically derive stroke-level structures from character bitmaps without handcrafted data, allowing for cross-lingual generalization.

Abstract

Hieroglyphs, as logographic writing systems, encode rich semantic and cultural information within their internal structural composition. Yet, current advanced Large Language Models (LLMs) and Multimodal LLMs (MLLMs) usually remain structurally blind to this information. LLMs process characters as textual tokens, while MLLMs additionally view them as raw pixel grids. Both fall short to model the underlying logic of character strokes. Furthermore, existing structural analysis methods are often script-specific and labor-intensive. In this paper, we propose Hieroglyphic Stroke Analyzer (HieroSA), a novel and generalizable framework that enables MLLMs to automatically derive stroke-level structures from character bitmaps without handcrafted data. It transforms modern logographic and ancient hieroglyphs character images into explicit, interpretable line-segment representations in a normalized coordinate space, allowing for cross-lingual generalization. Extensive experiments demonstrate that HieroSA effectively captures character-internal structures and semantics, bypassing the need for language-specific priors. Experimental results highlight the potential of our work as a graphematics analysis tool for a deeper understanding of hieroglyphic scripts. View our code at https://github.com/THUNLP-MT/HieroSA.

Enabling Stroke-Level Structural Analysis of Hieroglyphic Scripts without Language-Specific Priors

TL;DR

Hieroglyphic Stroke Analyzer (HieroSA) is proposed, a novel and generalizable framework that enables MLLMs to automatically derive stroke-level structures from character bitmaps without handcrafted data, allowing for cross-lingual generalization.

Abstract

Hieroglyphs, as logographic writing systems, encode rich semantic and cultural information within their internal structural composition. Yet, current advanced Large Language Models (LLMs) and Multimodal LLMs (MLLMs) usually remain structurally blind to this information. LLMs process characters as textual tokens, while MLLMs additionally view them as raw pixel grids. Both fall short to model the underlying logic of character strokes. Furthermore, existing structural analysis methods are often script-specific and labor-intensive. In this paper, we propose Hieroglyphic Stroke Analyzer (HieroSA), a novel and generalizable framework that enables MLLMs to automatically derive stroke-level structures from character bitmaps without handcrafted data. It transforms modern logographic and ancient hieroglyphs character images into explicit, interpretable line-segment representations in a normalized coordinate space, allowing for cross-lingual generalization. Extensive experiments demonstrate that HieroSA effectively captures character-internal structures and semantics, bypassing the need for language-specific priors. Experimental results highlight the potential of our work as a graphematics analysis tool for a deeper understanding of hieroglyphic scripts. View our code at https://github.com/THUNLP-MT/HieroSA.
Paper Structure (44 sections, 24 equations, 2 figures, 12 tables)

This paper contains 44 sections, 24 equations, 2 figures, 12 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Stroke Representation
  4. Reward Design for Stroke Representation
  5. Valid Stroke Identification
  6. Single-Stroke Coverage Estimation
  7. Reward Aggregation
  8. Training Paradigm
  9. Experiments
  10. Implementations
  11. Main Results
  12. Quantitative results.
  13. Qualitative results.
  14. Performance differences across training languages.
  15. Results on Downstream Tasks
  16. ...and 29 more sections

Figures (2)

  • Figure 1: Given a bitmap image of a hieroglyphic character (for example, the Oracle Bone Script character on the left), our proposed HieroSA infers stroke-level structure and converts the image into explicit line-segment representations in a normalized coordinate space (right).
  • Figure 2: Overview of HieroSA. (a) HieroSA takes a binarized character image as input and outputs a structured stroke representation, where strokes are represented as a set of line segments in a normalized coordinate space. (b) Illustration of the training objective: the predicted stroke segments are optimized to maximize their overlap with the black pixels in the binarized character image. (c) Geometric estimation of black-pixel coverage for a single stroke.