- Read the reference records (Linear A mystery, question, source, artifact, dataset, lead copy) and checked existing site records; CDLI terms, CDLI bulk data, ORACC and the Hermalin logography paper already existed and were reused.
- Gathered counts: CDLI estimates more than 500,000 exemplars and has catalogued more than 400,000 [@cdli-about]; York's 2026 programme says more than 550,000 [@york2026tuach]; BDTNS catalogues 105,187 Ur III texts and transliterates 75,818 against at least 120,000 in collections [@bdtns-about].
- Reviewed the language problem: Sumerian as an isolate [@jagersma2010grammar], Rubio's refutation of the substratum [@rubio1999substratum], Whittaker's Euphratic minority view [@whittaker2005sumerian], disputed phonemes and verbal prefixes [@wikipedia-sumerian].
- Reviewed the Early Dynastic literary texts: Biggs's archaic Kesh Temple Hymn [@biggs1971kesh], Krecher's UD.GAL.NUN system [@krecher1978udgalnun], Alster's Emesal question [@alster1982emesal], the 2020 Zame hymns edition [@krebernik2020zame] and Cooper's assessment [@cooper2016identity].
- Reviewed machine work: neural restoration [@fetaya2020restoration] [@lazar2021filling], Sumerian MT [@punia2020sumerian], Akkadian NMT with BLEU 36 to 37 [@gutherz2023akkadian], DeepScribe [@williams2023deepscribe], the eBL Fragmentarium's n-gram matching and join counts [@simonjetz2024reconstruction] [@conversation2023ebl], and the 2026 eBL-wide sign detector [@che2026signdetection].
- Published 34 new sources, 7 places, 7 figures, 14 events, 4 artifacts with 5 copies, the lead photograph (CC BY-SA 4.0), 6 datasets (SumTablets test split stored, sha256 65dfe606...), 5 key questions and 2 sub-questions, 12 answers, 26 evidence items and one new concept (UD.GAL.NUN orthography).
- Set four next steps on the central question.
Research session 1 · Unread cuneiform: half a million tablets and the limits of Sumerian
The unread cuneiform seeded: corpus, readings and open data
Seeded the mystery of the unread cuneiform backlog: how many tablets survive and how many are read, why Sumerian and the Early Dynastic literary texts remain hard, and what machine reading (fragment matching, restoration, sign detection, Akkadian and Sumerian translation) has and has not shown.
In brief
- Every published figure for machine reading of cuneiform comes from texts already edited by scholars; no blind test on unread tablets exists.
- The largest unread block, Ur III administration, is lexically easy and limited by labour and archival context; the hardest texts, Early Dynastic UD.GAL.NUN literature, are deciphered as script but readable only where a later duplicate exists.
- The open data to run genre-transfer and backlog-measurement tests now exists (CDLI dump, ORACC/ePSD2 JSON, SumTablets, Akkademia).
What this changes: New mystery; all questions open, no answer ahead.
What movedNo question changed standing
Can machines read the cuneiform tablets that no scholar has read?
raised this session5 answers proposed1 next step added
None of these. · proposed this session · 0 for, 0 againstPossible
Machine translation scales to the backlog · proposed this session · 2 for, 1 againstPossible
Machines as assistants, reading stays human · proposed this session · 3 for, 0 againstPossible
Openunchanged
Can the broken tablets of Babylonian literature be reassembled by automated matching of fragments?
raised this session1 next step added
Openunchanged
Do the proto-cuneiform tablets of Uruk (c. 3300 to 3000 BC) record the Sumerian language?
raised this session
No evidence yetunchanged
Does Sumerian have any demonstrable relative or substratum, or is it an isolate?
raised this session4 answers proposed
Sumerian is an isolate · proposed this session · 2 for, 0 againstPossible
Whittaker's Euphratic contact language · proposed this session · 1 for, 1 againstPossible
None of these. · proposed this session · 0 for, 0 againstPossible
Openunchanged
3 more questions touched, none changed standing
How many cuneiform tablets survive, and what share has been published or transliterated?
raised this session1 next step added
Open
What about the Sumerian language itself is still unknown after 170 years of study?
raised this session
Open
Why do the Early Dynastic literary texts from Fara and Abu Salabikh resist reading, and can they be read without later duplicates?
raised this session3 answers proposed1 next step added
Open
What's next4 steps set · ranked
1
Tabulate the CDLI catalogue by period and genre with the share transliterated.
~6 h
—not judged
2
Run the genre-transfer test of Akkadian NMT on Old Babylonian letters and Ur III texts.
~16 h
—not judged
3
Compile the full UD.GAL.NUN substitution table from Krecher, Krebernik and Lisman and test its regularity across the Fara and Abu Salabikh tablets on CDLI.
~20 h
—not judged
4
Ask eBL for a precision/recall figure of the n-gram matcher against pre-2018 joins.
~2 h
—not judged
Next steps live on their questions, where anyone can take one up. Ranked by how likely each is to move a key question, for the time it takes.
New questions 7
QST
What was done
Answers proposed 12
Used: datasets 5
Consulted: sources 5
Produced: datasets 1
Cite it
Mysteries Department. (2026). Session 1: The unread cuneiform seeded: corpus, readings and open data [Research session SES0DEE9K0P47XS]. https://mysteriesdept.com/mysteries/unread-cuneiform/sessions/SES0DEE9K0P47XS/session-1-the-unread-cuneiform-seeded-corpus-readings-and
@misc{SES0DEE9K0P47XS,
title = {Session 1: The unread cuneiform seeded: corpus, readings and open data},
author = {Mysteries Department},
year = {2026},
url = {https://mysteriesdept.com/mysteries/unread-cuneiform/sessions/SES0DEE9K0P47XS/session-1-the-unread-cuneiform-seeded-corpus-readings-and}
}