{"id":28753223,"url":"https://github.com/google-deepmind/ithaca","last_synced_at":"2025-10-05T09:44:31.362Z","repository":{"id":41039964,"uuid":"432137963","full_name":"google-deepmind/ithaca","owner":"google-deepmind","description":"Restoring and attributing ancient texts using deep neural networks","archived":false,"fork":false,"pushed_at":"2023-11-21T18:56:36.000Z","size":2500,"stargazers_count":559,"open_issues_count":1,"forks_count":62,"subscribers_count":16,"default_branch":"main","last_synced_at":"2025-06-14T01:46:58.313Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":"","language":"Jupyter Notebook","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/google-deepmind.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":"AUTHORS","dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2021-11-26T10:25:00.000Z","updated_at":"2025-06-03T12:11:23.000Z","dependencies_parsed_at":"2023-09-07T20:34:50.890Z","dependency_job_id":"d06abede-8ba9-4d24-8634-6c34af598c66","html_url":"https://github.com/google-deepmind/ithaca","commit_stats":null,"previous_names":["google-deepmind/ithaca","deepmind/ithaca"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/google-deepmind/ithaca","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/google-deepmind%2Fithaca","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/google-deepmind%2Fithaca/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/google-deepmind%2Fithaca/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/google-deepmind%2Fithaca/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/google-deepmind","download_url":"https://codeload.github.com/google-deepmind/ithaca/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/google-deepmind%2Fithaca/sbom","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":260268635,"owners_count":22983601,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2025-06-17T00:39:06.179Z","updated_at":"2025-10-05T09:44:26.302Z","avatar_url":"https://github.com/google-deepmind.png","language":"Jupyter Notebook","funding_links":[],"categories":[],"sub_categories":[],"readme":"\u003cp align=\"center\" style=\"margin-bottom:1em;\" \u003e\n\u003cimg alt=\"Ithaca logo\" src=\"images/ithaca-logo.svg\" width=\"33%\" /\u003e\n\u003c/p\u003e\n\n# Restoring and attributing ancient texts using deep neural networks\n\nYannis Assael\u003csup\u003e1,\\*\u003c/sup\u003e, Thea Sommerschield\u003csup\u003e2,3,\\*\u003c/sup\u003e, Brendan Shillingford\u003csup\u003e1\u003c/sup\u003e, Mahyar Bordbar\u003csup\u003e1\u003c/sup\u003e, John Pavlopoulos\u003csup\u003e4\u003c/sup\u003e,\nMarita Chatzipanagiotou\u003csup\u003e4\u003c/sup\u003e, Ion Androutsopoulos\u003csup\u003e4\u003c/sup\u003e, Jonathan Prag\u003csup\u003e3\u003c/sup\u003e, Nando de Freitas\u003csup\u003e1\u003c/sup\u003e\n\n*\u003csup\u003e1\u003c/sup\u003e DeepMind, United Kingdom\u003cbr/\u003e\n\u003csup\u003e2\u003c/sup\u003e Ca’ Foscari University of Venice, Italy\u003cbr/\u003e\n\u003csup\u003e3\u003c/sup\u003e University of Oxford, United Kingdom\u003cbr/\u003e\n\u003csup\u003e4\u003c/sup\u003e Athens University of Economics and Business, Greece\u003cbr/\u003e\n\u003csup\u003e\\*\u003c/sup\u003e Authors contributed equally to this work*\n\n---\n\n\u003ca href=\"https://colab.research.google.com/github/deepmind/ithaca/blob/master/colabs/ithaca_inference.ipynb\" target=\"_parent\"\u003e\u003cimg src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/\u003e\u003c/a\u003e\n\nAncient History relies on disciplines such as Epigraphy, the study of inscribed\ntexts known as \"inscriptions\", for evidence of the thought, language, society\nand history of past civilizations. However, over the centuries many inscriptions\nhave been damaged to the point of illegibility, transported far from their\noriginal location, and their date of writing is steeped in uncertainty. We\npresent Ithaca, the first Deep Neural Network for the textual restoration,\ngeographical and chronological attribution of ancient Greek inscriptions. Ithaca\nis designed to assist and expand the historian’s workflow: its architecture\nfocuses on collaboration, decision support, and interpretability.\n\n\u003cp align=\"center\" style=\"margin-top:2em; margin-bottom:2em;\" \u003e\n\u003cimg alt=\"Restoration of damaged inscription\" src=\"images/inscription.png\" width=\"75%\" /\u003e\u003cbr /\u003e\n\u003cem\u003eRestoration of damaged inscription: this inscription (\u003cit\u003eIG\u003c/it\u003e I\u003csup\u003e3\u003c/sup\u003e 4B) records a decree concerning the Acropolis of Athens and dates 485/4 BCE. (CC BY-SA 3.0, WikiMedia)\u003c/em\u003e\n\u003c/p\u003e\n\nWhile Ithaca alone achieves 62% accuracy when restoring damaged texts, as soon\nas historians use Ithaca their performance leaps from 25% to 72%, confirming\nthis synergistic research aid’s impact. Ithaca can attribute inscriptions to\ntheir original location with 71% accuracy and can date them with a distance of\nless than 30 years from ground-truth ranges, redating key texts of Classical\nAthens and contributing to topical debates in Ancient History. This work shows\nhow models like Ithaca can unlock the cooperative potential between AI and\nhistorians, transformationally impacting the way we study and write about one of\nthe most significant periods in human history.\n\n\u003cp align=\"center\" style=\"margin-top:2em; margin-bottom:2em;\" \u003e\n\u003cimg alt=\"Ithaca architecture\" src=\"images/ithaca-arch.png\" width=\"75%\" /\u003e\u003cbr /\u003e\n\u003cem\u003eIthaca's architecture processing the phrase \"δήμο το αθηναίων\" (\"the people of Athens\"). The first 3 characters of the phrase were hidden and their restoration is proposed. In tandem, Ithaca also predicts the inscription’s region and date.\u003c/em\u003e\n\u003c/p\u003e\n\n## References\n\n-   [Nature article](https://www.nature.com/articles/s41586-022-04448-z)\n-   [DeepMind blog](https://www.deepmind.com/blog/predicting-the-past-with-ithaca)\n\nWhen using any of this project's source code, please cite:\n\n```\n@article{asssome2022restoring,\n  title = {Restoring and attributing ancient texts using deep neural networks},\n  author = {Assael*, Yannis and Sommerschield*, Thea and Shillingford, Brendan and Bordbar, Mahyar and Pavlopoulos, John and Chatzipanagiotou, Marita and Androutsopoulos, Ion and Prag, Jonathan and de Freitas, Nando},\n  doi = {10.1038/s41586-022-04448-z},\n  journal = {Nature},\n  year = {2022}\n}\n```\n\n## Ithaca inference online\n\nTo aid further research in the field we created an online interactive python notebook, where researchers can query one of our trained models to get text restorations, visualise attention weights, and more.\n\n-   [Ithaca Interactive Interface](https://ithaca.deepmind.com/)\n-   [Google Colab for using Ithaca for your research](https://colab.research.google.com/github/deepmind/ithaca/blob/master/colabs/ithaca_inference.ipynb)\n\n## Ithaca inference offline\n\nAdvanced users who want to perform inference using the trained model may want\nto do so manually using the `ithaca` library directly.\n\nFirst, to install the `ithaca` library and its dependencies, run:\n```sh\npip install .\n```\n\nThen, download the model via\n```sh\ncurl --output checkpoint.pkl https://storage.googleapis.com/ithaca-resources/models/checkpoint_v1.pkl\n```\n\nAn example of using the library can be run via\n```sh\npython inference_example.py --input_file=example_input.txt\n```\nwhich will run restoration and attribution on\nthe text in `example_input.txt`.\n\nTo run it with different input text, run\n```sh\npython inference_example.py --input=\"...\"\n# or using text in a UTF-8 encoded text file:\npython inference_example.py --input_file=some_other_input_file.txt\n```\n\nThe restoration or attribution JSON can be saved to a file:\n```sh\npython inference_example.py \\\n  --input_file=example_input.txt \\\n  --attribute_json=attribute.json \\\n  --restore_json=restore.json\n```\n\nFor full help, run:\n```sh\npython inference_example.py --help\n```\n\n### Dataset generation\n\nIthaca was trained on The Packard Humanities Institute’s\n\"[Searchable Greek Inscriptions](https://inscriptions.packhum.org/)\" public\ndataset. The processing workflow for generating the machine-actionable text and\nmetadata, as well as further details on the train, validation and test splits\nare available at [I.PHI dataset](https://github.com/sommerschield/iphi).\n\n### Training Ithaca\nSee `train/README.md` for instructions.\n\n\n## License\nApache License, Version 2.0\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fgoogle-deepmind%2Fithaca","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fgoogle-deepmind%2Fithaca","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fgoogle-deepmind%2Fithaca/lists"}