Galen First Drafts: Machine First-Draft English Translations of Thirteen Works of the Galenic Corpus, with Their Checks

Much of Galen's surviving writing has never been translated into English in full. Galen First Drafts publishes machine first-draft English translations of thirteen works of the Galenic corpus (one of them, On Leeches, probably pseudo-Galenic) that have no complete English translation that we know of, from Kühn's Greek as digitized by First1KGreek (244,007 words of Greek). Every passage is shown beside its Greek, with two independent renderings by two language models of the same developer (Anthropic) and the outcome of three machine checks: back-translation into Greek, agreement between the two renderings on technical terms, and a dictionary-anchored model judge salted with planted errors and controls. The thresholds were fixed before each run; the judge's controls were rebuilt twice after its gate failed, and every verdict reported uses the last version. Passages that fail are marked and kept. In a reading of 39 random passages against the Greek by model readers, 4 had minor slips, all in passages the checks had passed. For two works the drafts were compared, term by term, with a scholar's English under a lenient rule (a target agrees if any content word of the rendering occurs anywhere in the scholar's passage): agreement with Manuela Marai's English of On the Composition of Drugs by Kind was 63.5% and 66.4% on ten calibration passages and 60.9% and 62.3% on 40 further passages from her thesis; with A. A. Long's English of one paragraph of On Difficulty of Breathing it was 2 and 3 of 14 terms. We also describe a catalogue of 678 items noted in the drafts by model readers. Two model research agents checked its items on lost authors and works and on Galen's own works against Fichtner's Corpus Galenicum (2023) and the fragment collections; none was shown to be unknown to scholarship. No scholar has read the drafts. They are for reading and for correction, not for quotation. Site: montblanctech.github.io/Galen/first-drafts/. Data and code: github.com/Montblanctech/Galen (concept DOI 10.5281/zenodo.21561962).

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-11
DOI
https://doi.org/10.5281/zenodo.24146102
Primary Topic
Translation Studies and Practices
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Galen First Drafts: Machine First-Draft English Translations of Thirteen Works of the Galenic Corpus, with Their Checks

William Pepperman
Zenodo (CERN European Organization for Nuclear Research)
Translation Studies and Practices
preprint

Galen First Drafts: Machine First-Draft English Translations of Thirteen Works of the Galenic Corpus, with Their Checks

William Pepperman
preprint en

Abstract

Much of Galen's surviving writing has never been translated into English in full. Galen First Drafts publishes machine first-draft English translations of thirteen works of the Galenic corpus (one of them, On Leeches, probably pseudo-Galenic) that have no complete English translation that we know of, from Kühn's Greek as digitized by First1KGreek (244,007 words of Greek). Every passage is shown beside its Greek, with two independent renderings by two language models of the same developer (Anthropic) and the outcome of three machine checks: back-translation into Greek, agreement between the two renderings on technical terms, and a dictionary-anchored model judge salted with planted errors and controls. The thresholds were fixed before each run; the judge's controls were rebuilt twice after its gate failed, and every verdict reported uses the last version. Passages that fail are marked and kept. In a reading of 39 random passages against the Greek by model readers, 4 had minor slips, all in passages the checks had passed. For two works the drafts were compared, term by term, with a scholar's English under a lenient rule (a target agrees if any content word of the rendering occurs anywhere in the scholar's passage): agreement with Manuela Marai's English of On the Composition of Drugs by Kind was 63.5% and 66.4% on ten calibration passages and 60.9% and 62.3% on 40 further passages from her thesis; with A. A. Long's English of one paragraph of On Difficulty of Breathing it was 2 and 3 of 14 terms. We also describe a catalogue of 678 items noted in the drafts by model readers. Two model research agents checked its items on lost authors and works and on Galen's own works against Fichtner's Corpus Galenicum (2023) and the fragment collections; none was shown to be unknown to scholarship. No scholar has read the drafts. They are for reading and for correction, not for quotation. Site: montblanctech.github.io/Galen/first-drafts/. Data and code: github.com/Montblanctech/Galen (concept DOI 10.5281/zenodo.21561962).

Zenodo (CERN European Organization for Nuclear Research)
Translation Studies and Practices
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon