Ithaca is a deep neural network system designed to assist historians in restoring, geographically attributing, and chronologically dating ancient Greek inscriptions. It targets epigraphers and ancient historians, offering a collaborative tool that enhances human expertise with AI-driven insights, improving accuracy and aiding scholarly debate.

How It Works

Ithaca employs a deep neural network architecture focused on collaboration, decision support, and interpretability. It processes damaged inscriptions to predict missing characters, suggest geographical origins, and estimate the period of writing. The system's design prioritizes synergy with human experts, aiming to augment rather than replace their knowledge.

Quick Start & Requirements

• Install via pip: pip install .
• Download model: curl --output checkpoint.pkl https://storage.googleapis.com/ithaca-resources/models/checkpoint_v1.pkl
• Run inference: python inference_example.py --input_file=example_input.txt
• Requires Python.
• An online interactive Colab notebook is available for easier use: Ithaca Interactive Interface

Highlighted Details

• Achieves 62% accuracy in text restoration, with human-AI collaboration boosting performance to 72%.
• Attributes inscriptions to their original location with 71% accuracy.
• Dates inscriptions with a median error of less than 30 years.
• Trained on the Packard Humanities Institute’s "Searchable Greek Inscriptions" dataset.

Maintenance & Community

The project is from Google DeepMind. Citation details for the Nature article are provided.

Licensing & Compatibility

• License: Apache License, Version 2.0.
• Permissive license suitable for commercial use and integration into closed-source projects.

Limitations & Caveats

The README does not detail specific hardware requirements beyond standard Python environments, nor does it mention any alpha/beta status or known limitations of the model's performance on specific types of inscriptions or historical periods.