Repository path: workshop/translations/hafez-bekonad/dependence-note.md · rendered 2026-09-09
Page metadata (front matter)
| type | note |
|---|---|
| id | dep-hafez-bekonad |
| status | frozen |
| created | 2026-08-30 |
| updated | 2026-08-30 |
| links | workshop/translations/hafez-bekonad/R57-v1/translation.md, runs/RS-20260830-leaf-contract/dep2.json, runs/RS-20260830-leaf-contract/dep.json, wiki/method-notes.md |
Contamination measurement — T-hafez-bekonad-R57-v1
Run with tools/dependence_check.py on both frozen arms, after the freeze and before the audit
was run. Raw output: runs/RS-20260830-leaf-contract/dep2.json (whole books) and dep.json
(odes only, and the two lead-versus-lead cells).
This measurement is stronger than the project's usual one, and the reason is worth stating. The standing rule's weak case is a comparator that does not contain the poem. Here it does: Leaf 1898 translated this very ghazal as his ode XV, and the comparator below is his complete book, ode XV included. The lead did not read that page before the freeze; the file was in the repository and the extraction that reads it was run afterwards.
| pair | shared 7-grams | 12-grams | 15-grams | longest run | verdict |
|---|---|---|---|---|---|
LEAF arm ~ Leaf 1898, the entire book (14,416 tokens, containing his rendering of this same ghazal) |
0 | 0 | 0 | 3 | clean |
FREE arm ~ Leaf 1898, the entire book |
0 | 0 | 0 | 3 | clean |
LEAF arm ~ Payne 1901 vol. 1, entire (71,268 tokens) |
0 | 0 | 0 | 4 | clean |
LEAF arm ~ the lead's own S232 rendering of sh118/sh124 |
0 | 0 | 0 | 2 | clean |
LEAF arm ~ FREE arm (the two arms against each other) |
0 | 0 | 0 | 6 | — |
Declared contamination: none, and this is the basis. Zero shared 7-grams against a published
Victorian rendering of the same poem.
Note (bhb) is not triggered and its figure is not comparable. The lead-versus-lead row is on
different poems (2 tokens). The last row is the two arms of the same rendering against each other
and is reported for completeness, not as an independence measurement: they are the same hand on the
same poem by design, and 6 tokens — will do . one prayer at and the shared rhyme-plus-radif tails
— is the floor a shared radif imposes.
Non-ASCII was 1.6% of characters in the first run and the tokeniser deletes non-ASCII letters; the whole-book run above is the reported one and the figure is recorded here as note (brk) asks.