Stylistic classification of cuneiform signs using convolutional neural networks

Abstract

The classification of cuneiform signs according to stylistic criteria is a difficult task, which often leaves experts in the field disagree. This study introduces a new publicly available dataset of cuneiform signs classified according to style and Convolutional Neural Network (CNN) approaches to differentiate between cuneiform signs of the two main styles of the first millennium bce, Neo-Assyrian and Neo-Babylonian. The CNN model reaches an accuracy of 83 % in style classification. This tool has potential implications for the recognition of individual scribes and the dating of undated cuneiform tablets.