Learning Diachronic Representations of Ancient Greek Letterforms

2026-06-23

AI Summary

This study addresses the challenges posed by the dramatic evolution of Ancient Greek handwritten glyphs over centuries—namely symbol variation, data scarcity, and textual degradation—by introducing three cross-era character datasets and a novel methodology combining similarity-weighted supervised contrastive loss with degradation-driven, domain-aware data augmentation. Leveraging both a lightweight CNN and a pretrained ResNet, the approach learns temporally consistent character embeddings while dynamically estimating inter-class similarity to guide representation learning. The proposed method significantly outperforms baseline models in recognition tasks, yielding embeddings that separate character classes more distinctly than those produced by PCA or generic pretrained models. These embeddings further enable the discovery of stylistic subgroups and facilitate diachronic visualization of glyph evolution, demonstrating both robustness and interpretability.

Abstract

Learning representations that remain robust across centuries of variation in handwriting is a key challenge in diachronic representation learning. Taking one of the longest continuously used writing systems, ancient Greek, as a case study, we introduce three datasets for diachronic representation learning: Hell-Char, a curated training set spanning the 3rd-1st centuries BCE, and two evaluation sets, PaLit-Char (2nd-5th c. CE) and Med-Char (9th-14th c. CE). To address the challenges of symbolic variation, scarce data, and systematic degradation, we propose: a similarity-weighted supervised contrastive loss that biases embeddings using dynamically estimated inter-class similarities, and a lacuna-driven augmentation scheme that simulates realistic manuscript corruptions. Trained with these strategies, both a lightweight CNN and a pretrained ResNet achieve strong recognition performance and produce embeddings that more coherently separate character classes than PCA or generic pretrained models. These embeddings enable clustering, identification of stylistic subgroups, and construction of prototype images that visualize diachronic evolution and transitional letterforms. Our results demonstrate that respecting intrinsic inter-letter relationships and augmenting with domain-informed corruptions yield robust, interpretable representations, offering a transferable paradigm for representation learning under scarce, temporally evolving, and noisy conditions. Code and data available at: https://github.com/ipavlopoulos/diachronic-greek-letterforms.

Problem

• diachronic representation learning
• ancient Greek letterforms
• handwriting variation
• scarce data
• manuscript degradation

Innovation

• diachronic representation learning
• supervised contrastive loss
• lacuna-driven augmentation
• ancient Greek letterforms
• handwriting variation