Code and Data for "Historical Insights from Sacrobosco Tables" Project

This repository contains code and data for the project:

"Insightful analysis of historical sources at scales beyond human capabilities using unsupervised Machine Learning and Explainable AI".

• sacrobosco_tables.zip: Contains the pre-processed Sacrobosco Tables Data. See folder raw/ for the original scans and processed/ for the binarized pages used as input for the model.

• page_data.zip: Contains three folders of training and evaluation material: patches_annotated/ are the ground annotations of digit patches and  contrast_annotated/ contains non-digit page content. The folder eval_annotated contains pages that have been annotated in full used for evaluation of full pages.

• code.zip: Contains the main code functionalities, including data processing, model training, inference and analysis and visualization code to generate case study analyses presented in our paper. See README.md for details on code execution and setup of the project repository.