Repository path: workshop/translations/saadi-ghazals-radif/dependence-note.md · rendered 2026-09-09

Contamination measurement — what was measured, and what it does not cover

Run 2026-08-26, after the translator's log was frozen and committed at 74cc7c83.

Comparator. Sir Lucas White King, The Odes of Sheikh Muslihud-Din Sa'di Shirazi (Kaviani Art Printing Press, Berlin, 1926), full text from archive.org/details/odesofsheikhsadishirazi. Published 1926 and in the public domain in the United States. It is the only freely reachable English of Sa'di's odes; every other candidate found was a Gulistan, which renders a different book.

Result (dependence.json): the 891 words of English in T-saadi-ghazals-radif-R51-v1 against the whole 68,000-word King volume —

And here is what that does not establish. King translates the بدایع, ordered by his own numbering; Ganjoor's «غزلیات» is a merged collection. Content probes run after the freeze — Shiraz beside a morning wind for ۴۹'s fourth bayt, cypress beside one foot for ۵۰'s fifth — do not locate King renderings of these six ghazals. The single "one foot" in the volume belongs to a different poem.

So this is a whole-corpus check and not a same-poem check. It says that 891 words of the lead's English of Sa'di share no seven-word run with 68,000 words of a published English of Sa'di in the same register — which is real evidence about diction, and is stronger than a same-poem check on a single short poem would be, because the comparator is seventy times the size. It is not evidence that the lead has not seen some other English of these particular poems.

The declaration therefore stays suspected and is not upgraded. Per the standing rule (CLAUDE.md, Contamination: measure before choosing): where no aligned comparator is reachable, say so on the artifact rather than let a measured-looking figure stand in for one.

Load-bearing? No. Every statistic this artifact feeds is a contrast between four arms written by the same hand at the same sitting, so nothing in the design turns on independence from any published rendering. The measurement is made because the rule requires it of every unit, not because a result depends on it.

Logged in wiki/base/consulted.md.