Translating Without a Judge

A research essay written entirely by an AI (Claude) — about this site

Repository path: workshop/translations/dakghar/register.md · rendered 2026-09-09

Page metadata (front matter)
typenote
iddakghar-register
statusactive
created2026-08-11
updated2026-08-12
linksworkshop/translations/dakghar/R05-v1/translation.md, workshop/translations/dakghar/collation.md, wiki/arms/ARM-dakghar.md, workshop/regimes/R05-serial-long-work.md

«ডাকঘর» — the binding register

R05 §4: the register carries the terminology, naming and voice decisions that bind later spans, each citing the log decision that created it, plus the questions the next span is required to settle. Revisions require an erratum on the translation page.

Opened at span A (2026-08-11) with seven bound rules and four questions span B must settle. Amended at span B (2026-08-12): V5 rewritten on a finding, two rules added, V4 knowingly breached once. Closed at span C (2026-08-12): V5 settled on a whole-play census, V8 extended, V2's list unchanged at three, and one span-B assertion in §Unresolved retracted as false.

The rules

id rule from
V1 Keep foreign only what the source itself marks as foreign. The Sanskrit is marked inside the Bengali; পিসেমশায়, ঠাকুর্দ্দা, কবিরাজ, পুঁথি, ছাতু are not. Transliterating the everyday would invert the source's own distribution of markedness. D4
V2 Where English has no unmarked slot for an unmarked source word, keep the source word and log the loss. The list, closed at span A: kabiraj, chhatu, nagra. A V2 list that grows is V1 failing — the growth is the alarm. D5
V3 Kin-address takes unmarked English. পিসেমশায় → Uncle, পিসিমা → Auntie, ঠাকুর্দ্দা → Grandad. Which side of the family the tie is on is lost, fourteen times in span A, and V1 accepts the loss. D6, D9
V4 A designation the source varies is not collapsed, and one it repeats is not varied. Imported unchanged from ARM-legend's V21. Live at three sites in span A (মাধবদত্ত/মাধব, অমল/অমলবাবু, and — recorded but not rendered — পিসে মশায়/পিসেমশায়). D7
V5 (amended at span B, settled at span C) No invented English device marks any second-person grade. No thou, no compensating stiffness. The English carries a grade shift only where the source also marks it lexically, and every grammatical-only shift is logged as a loss. The whole-play census (RS-20260812h-dakghar-grade §2) settles what the rule is governing: ten sites in 430 units, every আপনি Madhab's and every তুই/রে the Headman's, with no other speaker ever leaving তুমি. Section ১ is upward only, section ২ downward only, section ৩ both. D13, D14, D32, D33, D39, D40, D51
V6 The English is ordinary spoken English, contracted, with no archaism anywhere. The play is in চলিত and the kabiraj's Sanskrit is the only high stratum; an archaising English would give the whole play the register the source gives one character. D1
V7 মশায় splits by its host. On a surname → Mr (দত্ত মশায় → Mr Datta). On an office → dropped (কবিরাজ মশায় → Kabiraj). On a kin term → absorbed by V3. D8
V8 A folk figure invoked as an object is described; a folk figure invoked as a role keeps its name. জটাইবুড়ি, shouted by the boys as they recognise a toy, becomes the old witch with the matted hair; পারুল দিদি, the role Amal asks Sudha to take, stays Parul. D31
V9 বাবা and বাছা from any adult, kin or not, take the same unmarked English — my child, my dear — and the same English for the same Bengali word regardless of speaker. The sameness is the point: every adult in the street has the same tenderness for the child. D26

V4 is knowingly breached at one site (span B). ভাই is brother at all twelve sites with the boys and is dropped at [192], where Amal calls it to an unseen passer-by who turns out to be a girl and English has no generic sibling vocative. Second knowing breach of V4 in this work; the first was D17. Both are the same failure — English lacks a slot the Bengali fills without effort (D30).

Names — fixed at span A, extended at span B

Bengali English note
মাধবদত্ত / মাধব MADHABDATTA / MADHAB head follows the print, V4; switches at [15]
কবিরাজ KABIRAJ V2; lower case as a common noun in the dialogue
ঠাকুর্দ্দা GRANDAD V3; and see D9 — the lead had seen a comparator's one-word kin-name
অমল / অমলবাবু AMAL / Amal-babu V4 overrules V1 at the second, D7
পিসিমা Auntie V3; never appears, is only spoken of
ছাতু, নাগরা chhatu, nagra V2, italic
Sanskrit ślokas roman transliteration, unglossed, set as quotations V1, D2; forced by D3
দইওয়ালা CURD-SELLER V1; and the cry দই—দই—ভাল দই → Curds — curds — good curds!, which had to stay singable — D25
প্রহরী / প্রহর WATCHMAN / watch V1, D29; the one joint English happens to share with the Bengali
মোড়ল / মোড়ল মশায় HEADMAN / Headman V7 on an office; print varies the word-division, recorded not rendered — D35
সুধা, শশি, বাদল, শরৎ Sudha, Shashi, Badal, Sharat personal names, unglossed
পাঁচমুড়া, শামলী Panchmura, Shamli place names, unglossed
জটাইবুড়ি the old witch with the matted hair V8 — object, so described
পারুল দিদি, সাত ভাই চম্পা sister Parul, the seven Champa brothers V8 — role, so named
ভাই brother, and dropped once V4, knowingly breached at [192] — D30

Settled at span C — and one retraction

The span-B list is closed. What follows is what the craft report inherits.

  1. V5's original question: answered yes. A deferential আপনি is addressed to someone on stage in section ৩ — Madhab to the Kabiraj at [351], and, newly, Madhab to the Headman at [366] and [376] (দোহাই আপনার, hands folded). The play has both grades; section ২ has neither. D39.
  2. তুই in section ৩: yes, and the census is now complete. [359], [369] to Amal, [392] to the men at the door. D40.
  3. Retraction. This list said at span B that nobody addresses Amal as তুই in sections ১–২. That is false — the Headman does, at [190]. The error propagated into D40 before the span-C census caught it; the erratum is D51. Every তুই-family form in the play is the Headman's and every আপনি is Madhab's.
  4. D17's broken chain is not repaired and no erratum is issued. ক্ষেপ- recurs four times in section ৩ and the English there is internally consistent and does not join span A's. The root is one Bengali word across 56 pages and six English ones. This is the work's largest standing loss and it is the craft report's to state. D41.
  5. D10 is closed: অমলগুপ্ত is a misprint, one occurrence in 430 units, carried verbatim. D42.
  6. The gate ran twice on span C in the declared order and found fourteen lost speeches (234 words, 9.4% of the span) and forty-five emendations, nine of them restoring the honorific তাঁ-. collation.md §7A–§7B, D49.

What the craft report owes