Repository path: workshop/translations/hafez-boro/dependence-note.md · rendered 2026-09-09
Page metadata (front matter)
| type | note |
|---|---|
| id | hafez-boro-dependence |
| status | frozen |
| created | 2026-09-02 |
| updated | 2026-09-02 |
| links | workshop/translations/hafez-boro/R59-MONO-v1/translation.md, workshop/translations/hafez-boro/R59-PLAIN-v1/translation.md, wiki/method-notes.md |
Contamination measurement — T-hafez-boro-R59-MONO-v1 and -PLAIN-v1
Run after both logs were frozen and before the study limb was designed, per CLAUDE.md
§Contamination. Tool: python3 tools/dependence_check.py; raw output
runs/RS-20260902-rhyme-slot/dep.json.
Comparator: John Payne, The Poems of Shemseddin Mohammed Hafiz of Shiraz (1901), the
project's stored OCR at workshop/experiments/E-20260830-leaf-contract/materials/payne1901_djvu.txt
(394,713 characters). Payne's rendering of this ghazal is his ode 39, matched to Ganjoor sh35
at S236 (runs/RS-20260831b-radif-hands/match_accepted.json). Both the matched ode alone and the
whole stored corpus were used as comparators, because a longest run against one poem understates
what a corpus can supply.
| cell | shared 7-grams | 12 | 15 | longest common run | verdict |
|---|---|---|---|---|---|
MONO vs Payne ode 39 |
0 | 0 | 0 | 3 tokens | clean |
PLAIN vs Payne ode 39 |
0 | 0 | 0 | 3 tokens | clean |
MONO vs the whole stored Payne |
0 | 0 | 0 | 4 tokens | clean |
PLAIN vs the whole stored Payne |
0 | 0 | 0 | 4 tokens | clean |
MONO vs PLAIN (the two arms) |
39 | 13 | 6 | 20 tokens | expected, see below |
contamination: none on both arms. Zero shared seven-grams against a 394,713-character
comparator is the cleanest class of figure this project records, and it sits alongside
T-hafez-amad-R57-v1's (0 seven-grams, longest run 4) from S234.
The last row is not contamination; it is the design. R59 §4 requires the two arms to differ in
exactly one clause, so the six unconstrained lines are deliberately identical between them and the
20-token run is one of those lines. The tool has no way to know that, and its DEPENDENT? verdict on
a deliberately-paired cell is correct behaviour, not a finding.
What is not claimed. The stored Payne is OCR of one edition and the run length measured against it is a lower bound on what a longer or cleaner comparator would show. No claim is made about Leaf 1898, whose text was not used as a comparator here because he did not translate this ghazal.