Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Shaping History: Advanced Machine Learning Techniques for the Analysis and Dating of Cuneiform Tablets over Three Millennia

Danielle Kapon, Michael Fire, Shai Gordin

TL;DR

Shaping History demonstrates that tablet silhouettes encode substantial chronological information across more than 3,000 years. By combining a ResNet50-based silhouette classifier with VAEs trained on a 12-dimensional latent space, the work delivers an interpretable, scalable framework for dating cuneiform tablets and exploring shape evolution. Key contributions include large-scale shape analysis on 94,936 CDLI images, a 0.71 grayscale macro-F1 with ResNet50, and novel VAE-driven tools and widgets that visualize mean shapes and latent-factor trajectories across periods and genres. The approach offers a practical, data-driven complement to traditional diplomatics, enabling researchers to quantify shape-based patterns and standardize dating workflows in quantitative archaeology.

Abstract

Cuneiform tablets, emerging in ancient Mesopotamia around the late fourth millennium BCE, represent one of humanity's earliest writing systems. Characterized by wedge-shaped marks on clay tablets, these artifacts provided insight into Mesopotamian civilization across various domains. Traditionally, the analysis and dating of these tablets rely on subjective assessment of shape and writing style, leading to uncertainties in pinpointing their exact temporal origins. Recent advances in digitization have revolutionized the study of cuneiform by enhancing accessibility and analytical capabilities. Our research uniquely focuses on the silhouette of tablets as significant indicators of their historical periods, diverging from most studies that concentrate on textual content. Utilizing an unprecedented dataset of over 94,000 images from the Cuneiform Digital Library Initiative collection, we apply deep learning methods to classify cuneiform tablets, covering over 3,000 years of history. By leveraging statistical, computational techniques, and generative modeling through Variational Auto-Encoders (VAEs), we achieve substantial advancements in the automatic classification of these ancient documents, focusing on the tablets' silhouettes as key predictors. Our classification approach begins with a Decision Tree using height-to-width ratios and culminates with a ResNet50 model, achieving a 61% macro F1-score for tablet silhouettes. Moreover, we introduce novel VAE-powered tools to enhance explainability and enable researchers to explore changes in tablet shapes across different eras and genres. This research contributes to document analysis and diplomatics by demonstrating the value of large-scale data analysis combined with statistical methods. These insights offer valuable tools for historians and epigraphists, enriching our understanding of cuneiform tablets and the cultures that produced them.

Shaping History: Advanced Machine Learning Techniques for the Analysis and Dating of Cuneiform Tablets over Three Millennia

TL;DR

Shaping History demonstrates that tablet silhouettes encode substantial chronological information across more than 3,000 years. By combining a ResNet50-based silhouette classifier with VAEs trained on a 12-dimensional latent space, the work delivers an interpretable, scalable framework for dating cuneiform tablets and exploring shape evolution. Key contributions include large-scale shape analysis on 94,936 CDLI images, a 0.71 grayscale macro-F1 with ResNet50, and novel VAE-driven tools and widgets that visualize mean shapes and latent-factor trajectories across periods and genres. The approach offers a practical, data-driven complement to traditional diplomatics, enabling researchers to quantify shape-based patterns and standardize dating workflows in quantitative archaeology.

Abstract

Cuneiform tablets, emerging in ancient Mesopotamia around the late fourth millennium BCE, represent one of humanity's earliest writing systems. Characterized by wedge-shaped marks on clay tablets, these artifacts provided insight into Mesopotamian civilization across various domains. Traditionally, the analysis and dating of these tablets rely on subjective assessment of shape and writing style, leading to uncertainties in pinpointing their exact temporal origins. Recent advances in digitization have revolutionized the study of cuneiform by enhancing accessibility and analytical capabilities. Our research uniquely focuses on the silhouette of tablets as significant indicators of their historical periods, diverging from most studies that concentrate on textual content. Utilizing an unprecedented dataset of over 94,000 images from the Cuneiform Digital Library Initiative collection, we apply deep learning methods to classify cuneiform tablets, covering over 3,000 years of history. By leveraging statistical, computational techniques, and generative modeling through Variational Auto-Encoders (VAEs), we achieve substantial advancements in the automatic classification of these ancient documents, focusing on the tablets' silhouettes as key predictors. Our classification approach begins with a Decision Tree using height-to-width ratios and culminates with a ResNet50 model, achieving a 61% macro F1-score for tablet silhouettes. Moreover, we introduce novel VAE-powered tools to enhance explainability and enable researchers to explore changes in tablet shapes across different eras and genres. This research contributes to document analysis and diplomatics by demonstrating the value of large-scale data analysis combined with statistical methods. These insights offer valuable tools for historians and epigraphists, enriching our understanding of cuneiform tablets and the cultures that produced them.
Paper Structure (30 sections, 18 figures, 1 table)

This paper contains 30 sections, 18 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Tablet Dating
  4. Using Machine Learning for Tablet Dating
  5. Interpretability of Machine Learning and Deep Learning models
  6. Analysis by Synthesis
  7. The Generative Learning Trilemma
  8. Methods
  9. Data Acquisition
  10. Data Preprocessing
  11. Exploratory Data Analysis - Height-to-Width Ratios
  12. Tablet Period Classification
  13. Evaluation Metrics for Image-Based Period Classification
  14. Classic Machine Learning models
  15. Unsupervised Feature Extraction
  16. ...and 15 more sections

Figures (18)

  • Figure 1: A sample of some of the variety of shapes and sizes of cuneiform tablets texts: (a) Neo-Assyrian; (b) Ur III; (c) Middle Babylonian; (d) Neo-Babylonian.
  • Figure 2: Number of Samples per Period and Genre
  • Figure 3: Interactive widgets for exploring tablet characteristics and evolution.
  • Figure 4: Tablets addressed in the preprocessing stage of the research
  • Figure 5: Illustration of the process for extracting the largest component of a tablet.
  • ...and 13 more figures