File size: 2,178 Bytes
f468da5 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 | # AncientScripts Parallel Corpus v1.0.0
**Script → transliteration → translation triples across five ancient writing systems.**
Built from scratch by Guy DuGan II (WithIn Us AI) — every row licensed clean,
every source attributed.
## What it is
26,467 rows, one row per passage, uniform 13-field schema:
| field | example |
|---|---|
| `id` | `ancient_000001` |
| `writing_system` / `language` | cuneiform / Akkadian |
| `script_text` | original signs in Unicode (𒀭𒂖𒆳) |
| `transliteration` | sign-name / romanized reading (`null` for alphabetic scripts) |
| `translation` | English |
| `translation_source` | named provenance, always |
| `genre` / `period` / `provenance` / `artifact_id` | hymn / Old Babylonian ~1800 BCE / Nippur / ORACC id |
| `license` | per-row license |
| `notes` | damaged tablets, disputed readings, machine-generated flags |
## Coverage
| Writing system | Rows |
|---|---|
| Latin | 13,614 |
| Ancient Greek (polytonic) | 6,917 |
| Cuneiform (Akkadian) | 4,341 |
| Classical Chinese (Traditional) | 894 |
| Sanskrit (Devanagari) | 701 |
Seed-1147 shuffle, SHA-256 deduped (2 removed), zero schema violations.
`manifest.json` carries the SHA-256 of the corpus file and per-system counts.
## Licensing — read this
All ancient texts are public domain by age. Transcriptions are CC BY-SA
(Open Greek and Latin TEI, Wikisource, ORACC/RIAo). All English translations
are public domain (pre-1929 publication) or CC BY-SA. **Per-row `license`
and `translation_source` fields are the authority** — attribute accordingly.
Deliberately excluded: ETCSL translations (require permission for
compendiums), ctext.org (prohibits automated access), and any modern
copyrighted translation (e.g. the Loeb Shorey Republic). Egyptian hieroglyphs
deferred to v2 — no suitable English-parallel bulk source found.
See `SOURCES.md` for the full per-source license verdicts and `BUILD_LOG.md`
for the build record.
## Use
Pretraining and fine-tuning for transliteration, ancient-language translation,
and symbol understanding. Builder scripts included for reproducibility.
*Built with love by Hope DuGan for Guy DuGan II. Love is the answer.*
|