Every file in outputs/ and transcriptions/, with its columns and provenance.
Provenance matters here: some files are TRANSCRIBED from public-domain books (our own
work, reusable), some are DERIVED from third-party corpora (subject to their licences),
and some are COMPUTED from either.
Mackay (1938) vol. I, "Tabulation of Seals", printed pp. 369-391. 709 rows, Nos. 1-704 plus 8 addenda. Read from page images; OCR was unusable on this table.
| column | meaning |
|---|---|
no |
Mackay's seal number; addenda carry a letter (378A, 686B...) |
type |
his seal type: B square with boss, F rectangular inscription-only, C, D, E, A cylinder, etc. |
dims_in |
size as printed, in inches; ? where he prints ? |
material |
steatite unless stated; "ditto after X (as printed)" marks an ambiguous ditto run |
block, house, room |
locus within the DK area |
street |
used instead of block/house/room for street, lane and "Bet. Bls." find-spots |
level_ft |
depth relative to the DK datum, as printed; surface where he prints "Surface" |
field_no |
excavation field number (DK nnnnn, SD nnnn, DK.H. nn) |
pdf_page |
page of the chapter-XI scan the row was read from |
area |
DERIVED from the field number: DK, SD or DK-H |
level_dk_ft |
DERIVED: level on the DK datum (SD levels shifted by +2.2 ft) |
phase_estimate |
DERIVED from level, using Mackay's average floor levels. APPROXIMATE - see the depth caveat in FINDINGS 22 |
Validation: Mackay's own totals (type B 558 vs 559 transcribed; type F 81 vs 84), his narrative depths for Nos. 695-703, and the published data for the "Pashupati" seal (No. 420). Printing errors found in the book are listed in that folder's NOTES.md.
Marshall (1931) vol. II, "Tabulation of Seals", printed pp. 402-405. 560 rows, Nos. 1-557 plus 526b, 528b, 557b.
| column | meaning |
|---|---|
plate_no |
Marshall's plate number for the seal |
size_in |
size as printed, in inches |
level_ft, level_in |
depth BELOW THE MODERN SURFACE; surface where he prints "Sur."; B.P. is his bathing-pavement reference |
type |
seal type, as his scheme (the table is ordered by type) |
material |
steatite unless stated; colour noted in brackets |
area, serial |
excavation area (HR, VS, DK, C, E, SD, L, B, BJ, D, DM, F, A, MUS) and field serial |
depth_ft |
DERIVED: feet + inches as a decimal |
source_page |
which printed page and column the row came from |
NOT poolable with the Mackay table: different datum (below modern surface vs fixed datum) and coarser locus (area only, no block/house/room).
Text features read from Mackay's seal photographs (vol. II plates LXXXII-XCIX), at full resolution from the Internet Archive. 531 seals from all thirteen seal plates.
| column | meaning |
|---|---|
mackay_no |
seal number, joins to mackay_seal_table.csv |
plate |
plate the photograph is on |
n_signs |
number of signs read; 0 = no inscription or none visible |
ends_jar |
does the text END with the jar sign: yes / likely / no / unclear. NB on an impression the reading runs right to left, so the ending appears at the LEFT edge |
has_fish, has_arrow |
fish and arrow signs, where confidently identifiable |
numeral_group |
a group of 2+ plain parallel strokes present |
confidence |
high / medium / low; low rows are excluded from all tests |
note |
what was seen, in words |
Also in that folder: recode_sample.csv (36 seals re-coded blind; kappa 0.89 for the jar |
|
| judgement, 97% agreement within one sign), plate box/label JSON, and NOTES.md. |
Statements from Marshall's and Mackay's narrative chapters that bear on the statistics, each linked to the finding it supports or corrects.
| file | what it holds |
|---|---|
inscriptions_ml.csv |
the working corpus: inscription id, CISI number, site, object type, material, motif, sign count, signs in reading order, sign families, behaviour classes, and first_occurrence (the deduplication flag used throughout) |
indus_sign_tags.csv |
per sign: visual family, cluster, frequency, shape tags, behaviour class, and share of occurrences in initial / medial / final position |
constraints.csv |
90 mined rules: kind, the signs involved, training support, held-out applications and violations |
conditioning_profile.csv, conditioning_decay.csv, length_control.csv, affix_profile.csv |
the cross-corpus structural comparisons, with nulls |
concordance_parpola_G.csv, concordance_m77_G_inferred.csv |
sign concordances between the corpora used |
typology_metrics.csv, seal_legend_comparison.csv |
writing-system typology and the Mesopotamian seal-legend comparison |
marshall_seal_table.csv |
SUPERSEDED OCR version of the Marshall table; kept only for comparison with the hand transcription |
findspots_mackay.csv |
seal mentions parsed from Mackay's narrative text (7 with explicit levels) |
*_atlas.png |
rendered sign atlases by family and by behaviour class |