Source and licence
share: attributionnoneSource: corpus/incoming/hieratic-image-data/PaPYrus_Dataset-By-Sign.zip (staged 2026-09-10 by worker W-B2; provenance, terms as fetched and SHA-256 in corpus/incoming/hieratic-image-data/README.md)
Attribution: Julius Tabin, PaPYrus: Optical Character Recognition Applied to Ancient Egyptian Hieratic (University of Chicago NELC thesis, 2022), https://github.com/jtabin/PaPYrus, licensed GNU GPL v3.0. SHARE-ALIKE: a distributed derivative of this data must also be GPL-3.0, and the author asks that additions to the data set be released. The facsimile makers are named per image in the file names.
Licence: attribution — GNU General Public License v3.0, the licence Julius Tabin put on the PaPYrus data set (LICENSE.md in github.com/jtabin/PaPYrus, fetched 2026-09-10; GitHub spdx_id GPL-3.0). Copyleft: a distributed derivative of this data must itself be GPL-3.0, and the author asks in the README that additions to the data set be released. RI-036 RULED 2026-09-12 by Yousef (consistent with the 2026-09-10 ruling): a model trained on this data set IS released GPL-3.0, and training on staged material is allowed; the question of REPORT-W-B2 is closed
The text
10 lines · a line number is a link to itself| ref | L0 witness |
|---|---|
| 1 | THE FIRST HIERATIC IMAGE DATA IN THE PROJECT. 15,267 PNG crops named <Gardiner code>_<instance>_<facsimile/provenance tags>.png (ligatures keep both codes, e.g. N37_N35_0001_1_2_9.png), foldered three ways: by sign class, by provenance/text/facsimile maker, and flat. The facsimile-maker tag is a READY-MADE TRAIN/TEST SPLIT that tests generalisation across hands. Verified by extracting and rendering three crops: solid-black cursive silhouettes from facsimiles — D21 as the flattened hieratic stroke, M4 and S29 as tall cursive forms — hieratic, not hieroglyphic. A Hugging Face dataset search for "hieratic" returns ZERO results; there is no other hieratic dataset. From Julius Tabin's University of Chicago NELC thesis, From Papyrus to Pixels (2022), staged beside the data. |
| 2 | Staged files: |
| 3 | · corpus/incoming/hieratic-image-data/PaPYrus_Dataset-By-Sign.zip — 27,105,481 bytes |
| 4 | · corpus/incoming/hieratic-image-data/PaPYrus_Dataset-Whole.zip — 26,351,103 bytes |
| 5 | · corpus/incoming/hieratic-image-data/PaPYrus_Dataset-Categorized.zip — 27,023,072 bytes |
| 6 | · corpus/incoming/hieratic-image-data/PaPYrus_Precalculated-Data-Set-Stats.zip — 3,657,457 bytes |
| 7 | · corpus/incoming/hieratic-image-data/PaPYrus_Tabin-Thesis-2022.pdf — 259 pages, 38,984,165 bytes |
| 8 | · corpus/incoming/hieratic-image-data/PaPYrus_LICENSE.md — 35,149 bytes |
| 9 | · corpus/incoming/hieratic-image-data/PaPYrus_README.md — 5,126 bytes |
| 10 | THE PAGES ARE NOT IN THIS REPOSITORY. This document is a provenance stub: the original is staged under Coptic OCR/corpus/incoming/, with its SHA-256 and the terms as fetched in the README beside it. The catalogue row's `words` count is this stub's, not the volume's. |
Provenance
- title
- Tabin, PaPYrus: 15,267 hieratic sign crops labelled by Gardiner code (2022)
- dialect
- none
- source
- corpus/incoming/hieratic-image-data/PaPYrus_Dataset-By-Sign.zip (staged 2026-09-10 by worker W-B2; provenance, terms as fetched and SHA-256 in corpus/incoming/hieratic-image-data/README.md)
- licence
- attribution — GNU General Public License v3.0, the licence Julius Tabin put on the PaPYrus data set (LICENSE.md in github.com/jtabin/PaPYrus, fetched 2026-09-10; GitHub spdx_id GPL-3.0). Copyleft: a distributed derivative of this data must itself be GPL-3.0, and the author asks in the README that additions to the data set be released. RI-036 RULED 2026-09-12 by Yousef (consistent with the 2026-09-10 ruling): a model trained on this data set IS released GPL-3.0, and training on staged material is allowed; the question of REPORT-W-B2 is closed
- share
- attribution
- attribution
- Julius Tabin, PaPYrus: Optical Character Recognition Applied to Ancient Egyptian Hieratic (University of Chicago NELC thesis, 2022), https://github.com/jtabin/PaPYrus, licensed GNU GPL v3.0. SHARE-ALIKE: a distributed derivative of this data must also be GPL-3.0, and the author asks that additions to the data set be released. The facsimile makers are named per image in the file names.
- file
- Ancient Egyptian/Hieratic/PaPYrus — labelled hieratic sign images/Tabin, PaPYrus- 15,267 hieratic sign crops labelled by Gardiner code (2022).data.L0.txt
L0 witness
Cite as:
Tabin, PaPYrus: 15,267 hieratic sign crops labelled by Gardiner code (2022). corpus/incoming/hieratic-image-data/PaPYrus_Dataset-By-Sign.zip (staged 2026-09-10 by worker W-B2; provenance, terms as fetched and SHA-256 in corpus/incoming/hieratic-image-data/README.md). Layers as published by the Kemetic project (Yousef Hanna), 2026: Coptic · Cairene · Kemetic, /doc/ancient-egyptian/hieratic/papyrus-labelled-hieratic-sign-images/tabin-papyrus-15-267-hieratic-sign-crops-labelled-by-gardiner-code-2022.data/. Julius Tabin, PaPYrus: Optical Character Recognition Applied to Ancient Egyptian Hieratic (University of Chicago NELC thesis, 2022), https://github.com/jtabin/PaPYrus, licensed GNU GPL v3.0. SHARE-ALIKE: a distributed derivative of this data must also be GPL-3.0, and the author asks that additions to the data set be released. The facsimile makers are named per image in the file names.
What the layers are
The text as its witness or edition has it, original spelling; its dialect is a description and is never standardized.
For Egyptian texts: the transliteration with the editors' marks turned into the symbols ° * < > _ ^ (ruled 2026-09-10). For papyri: the edition's Leiden marks as the same symbols, and the editors' readings beside the scribe's as ‹scribe→editors› (ruled PAP-3).
Standardized Greco-Bohairic spelling and grammar, generated by code from the layer above under the rulings of the project.
The Kemetic alphabet, generated by code from L1.
Quoted translations where they exist, credited; otherwise model drafts, marked as drafts.