Restoring ancient text using deep learning: a case study on Greek epigraphy

Restoring ancient text using deep learning: a case study on Greek epigraphy. Assael, Y., Sommerschield, T., & Prag, J. October, 2019. arXiv:1910.06262 [cs]

Paper abstract bibtex

Ancient History relies on disciplines such as Epigraphy, the study of ancient inscribed texts, for evidence of the recorded past. However, these texts, “inscriptions”, are often damaged over the centuries, and illegible parts of the text must be restored by specialists, known as epigraphists. This work presents PYTHIA, the ﬁrst ancient text restoration model that recovers missing characters from a damaged text input using deep neural networks. Its architecture is carefully designed to handle longterm context information, and deal efﬁciently with missing or corrupted character and word representations. To train it, we wrote a nontrivial pipeline to convert PHI, the largest digital corpus of ancient Greek inscriptions, to machine actionable text, which we call PHI-ML. On PHI-ML, PYTHIA’s predictions achieve a 30.1% character error rate, compared to the 57.3% of human epigraphists. Moreover, in 73.5% of cases the ground-truth sequence was among the Top-20 hypotheses of PYTHIA, which effectively demonstrates the impact of this assistive method on the ﬁeld of digital epigraphy, and sets the state-of-the-art in ancient text restoration.

@misc{assael_restoring_2019,
	title = {Restoring ancient text using deep learning: a case study on {Greek} epigraphy},
	shorttitle = {Restoring ancient text using deep learning},
	url = {http://arxiv.org/abs/1910.06262},
	abstract = {Ancient History relies on disciplines such as Epigraphy, the study of ancient inscribed texts, for evidence of the recorded past. However, these texts, “inscriptions”, are often damaged over the centuries, and illegible parts of the text must be restored by specialists, known as epigraphists. This work presents PYTHIA, the ﬁrst ancient text restoration model that recovers missing characters from a damaged text input using deep neural networks. Its architecture is carefully designed to handle longterm context information, and deal efﬁciently with missing or corrupted character and word representations. To train it, we wrote a nontrivial pipeline to convert PHI, the largest digital corpus of ancient Greek inscriptions, to machine actionable text, which we call PHI-ML. On PHI-ML, PYTHIA’s predictions achieve a 30.1\% character error rate, compared to the 57.3\% of human epigraphists. Moreover, in 73.5\% of cases the ground-truth sequence was among the Top-20 hypotheses of PYTHIA, which effectively demonstrates the impact of this assistive method on the ﬁeld of digital epigraphy, and sets the state-of-the-art in ancient text restoration.},
	language = {en},
	urldate = {2023-01-26},
	publisher = {arXiv},
	author = {Assael, Yannis and Sommerschield, Thea and Prag, Jonathan},
	month = oct,
	year = {2019},
	note = {arXiv:1910.06262 [cs]},
	keywords = {Computer Science - Computation and Language, Computer Science - Computers and Society},
}

Downloads: 0

{"_id":"rhevQ2e6khwd2KPTJ","bibbaseid":"assael-sommerschield-prag-restoringancienttextusingdeeplearningacasestudyongreekepigraphy-2019","author_short":["Assael, Y.","Sommerschield, T.","Prag, J."],"bibdata":{"bibtype":"misc","type":"misc","title":"Restoring ancient text using deep learning: a case study on Greek epigraphy","shorttitle":"Restoring ancient text using deep learning","url":"http://arxiv.org/abs/1910.06262","abstract":"Ancient History relies on disciplines such as Epigraphy, the study of ancient inscribed texts, for evidence of the recorded past. However, these texts, “inscriptions”, are often damaged over the centuries, and illegible parts of the text must be restored by specialists, known as epigraphists. This work presents PYTHIA, the ﬁrst ancient text restoration model that recovers missing characters from a damaged text input using deep neural networks. Its architecture is carefully designed to handle longterm context information, and deal efﬁciently with missing or corrupted character and word representations. To train it, we wrote a nontrivial pipeline to convert PHI, the largest digital corpus of ancient Greek inscriptions, to machine actionable text, which we call PHI-ML. On PHI-ML, PYTHIA’s predictions achieve a 30.1% character error rate, compared to the 57.3% of human epigraphists. Moreover, in 73.5% of cases the ground-truth sequence was among the Top-20 hypotheses of PYTHIA, which effectively demonstrates the impact of this assistive method on the ﬁeld of digital epigraphy, and sets the state-of-the-art in ancient text restoration.","language":"en","urldate":"2023-01-26","publisher":"arXiv","author":[{"propositions":[],"lastnames":["Assael"],"firstnames":["Yannis"],"suffixes":[]},{"propositions":[],"lastnames":["Sommerschield"],"firstnames":["Thea"],"suffixes":[]},{"propositions":[],"lastnames":["Prag"],"firstnames":["Jonathan"],"suffixes":[]}],"month":"October","year":"2019","note":"arXiv:1910.06262 [cs]","keywords":"Computer Science - Computation and Language, Computer Science - Computers and Society","bibtex":"@misc{assael_restoring_2019,\n\ttitle = {Restoring ancient text using deep learning: a case study on {Greek} epigraphy},\n\tshorttitle = {Restoring ancient text using deep learning},\n\turl = {http://arxiv.org/abs/1910.06262},\n\tabstract = {Ancient History relies on disciplines such as Epigraphy, the study of ancient inscribed texts, for evidence of the recorded past. However, these texts, “inscriptions”, are often damaged over the centuries, and illegible parts of the text must be restored by specialists, known as epigraphists. This work presents PYTHIA, the ﬁrst ancient text restoration model that recovers missing characters from a damaged text input using deep neural networks. Its architecture is carefully designed to handle longterm context information, and deal efﬁciently with missing or corrupted character and word representations. To train it, we wrote a nontrivial pipeline to convert PHI, the largest digital corpus of ancient Greek inscriptions, to machine actionable text, which we call PHI-ML. On PHI-ML, PYTHIA’s predictions achieve a 30.1\\% character error rate, compared to the 57.3\\% of human epigraphists. Moreover, in 73.5\\% of cases the ground-truth sequence was among the Top-20 hypotheses of PYTHIA, which effectively demonstrates the impact of this assistive method on the ﬁeld of digital epigraphy, and sets the state-of-the-art in ancient text restoration.},\n\tlanguage = {en},\n\turldate = {2023-01-26},\n\tpublisher = {arXiv},\n\tauthor = {Assael, Yannis and Sommerschield, Thea and Prag, Jonathan},\n\tmonth = oct,\n\tyear = {2019},\n\tnote = {arXiv:1910.06262 [cs]},\n\tkeywords = {Computer Science - Computation and Language, Computer Science - Computers and Society},\n}\n\n\n\n","author_short":["Assael, Y.","Sommerschield, T.","Prag, J."],"key":"assael_restoring_2019","id":"assael_restoring_2019","bibbaseid":"assael-sommerschield-prag-restoringancienttextusingdeeplearningacasestudyongreekepigraphy-2019","role":"author","urls":{"Paper":"http://arxiv.org/abs/1910.06262"},"keyword":["Computer Science - Computation and Language","Computer Science - Computers and Society"],"metadata":{"authorlinks":{}}},"bibtype":"misc","biburl":"https://bibbase.org/zotero-group/schulzkx/5158478","dataSources":["JFDnASMkoQCjjGL8E"],"keywords":["computer science - computation and language","computer science - computers and society"],"search_terms":["restoring","ancient","text","using","deep","learning","case","study","greek","epigraphy","assael","sommerschield","prag"],"title":"Restoring ancient text using deep learning: a case study on Greek epigraphy","year":2019}