Executive Summary: Ezekiel Bengali Translation Requirements

Why it matters

Ezekiel is the most theologically dense, terminologically hazardous book yet processed for the Bengali Bible-translation pipeline. Its core passage — the Valley of Dry Bones (37:1-14) — sits directly on top of Bengal’s single most fraught vocabulary collision (resurrection vs. reincarnation), while the book’s 217 occurrences of the compound divine title “Lord GOD,” 120 occurrences of the messenger formula, and 60+ occurrences of the recognition formula demand a level of fixed-phrase discipline with no precedent in the Romans baseline. Get the glossary wrong here and Bengali readers — drawn from West Bengal’s Hindu-majority devotional culture and Bangladesh’s Muslim-majority context alike — will hear either reincarnation, ritual idol-invocation, or a biologically literal divine sonship where the text intends none of those things.

Key findings

• 20 doctrines analyzed, full-book coverage (chs. 1–48), extending and never contradicting the Romans baseline: 7 Critical, 10 High, 3 Medium, 0 Low — 17 require mandatory human theologian review, 3 require native-speaker review, none are automated-only.
• Glossary contains 72 total terms: 16 inherited exactly from the Romans baseline (5 Critical, 5 High, 4 Medium, 2 Low) plus 56 new Ezekiel-specific terms (14 Critical, 23 High, 19 Medium) — 19 Critical-risk and 28 High-risk terms in total, all routed to mandatory theologian review.
• The book’s central plot (glory departing chs. 8-11, returning ch. 43-44) and its two governing liturgical refrains (recognition formula, messenger formula) require string-match-level consistency enforcement beyond ordinary doctrinal review, given their combined ~400 occurrences.
• The core passage’s fourfold wordplay on רוּחַ (breath/wind/Spirit/“my Spirit”) cannot be carried by any single Bengali word — this is documented as the single highest-ranked unresolved translation ambiguity in the entire book.

Risks

• Idolatry vocabulary (প্রতিমা, ~40+ occurrences): the same word Bengali Hindus use reverently for Durga Puja images; requires standing pastoral teaching notes at every major cluster, not euphemistic substitution — the single most acute pastoral-sensitivity risk in the curriculum.
• “Soul” in Ezekiel 18 (נֶפֶש): if rendered আত্মা instead of প্রাণ, Bengali readers will hear an indestructible Vedantic ātman — the exact opposite of the chapter’s doctrine that the whole mortal person dies for sin.
• Dry-bones imagery (37:1-14): any lexical or framing slip risks readers hearing পুনর্জন্ম (reincarnation) rather than the intended corporate-restoration imagery that only typologically anticipates বodily resurrection.
• “Son of man” address (93 occurrences): unmarked reuse of মনুষ্যপুত্র (Christ’s fixed NT title) would import divine-authority weight backward into God’s address stressing Ezekiel’s creaturely lowliness — a direct doctrinal inversion.
• Sea-monster imagery for Pharaoh: any naga/serpent-deity-adjacent vocabulary would collide with Bengal’s still-living Manasa Devi folk cult.
• Temple vocabulary (মন্দির): must always carry a divine-name qualifier (সদাপ্রভুর মন্দির) to avoid both generic Hindu-temple association and confusion with the baseline’s rejected-for-church convention.

Opportunities

• Ezekiel’s watchman doctrine (প্রহরী) and its personal-accountability logic translate unusually cleanly into the Bengali Muslim conceptual world via the Qur’anic warner (nadhir) parallel — a strong teaching bridge requiring no vocabulary compromise.
• The river-of-living-water and tree-of-healing imagery (ch. 47) offers a ready-made Eden→Ezekiel→Revelation typological teaching chain that native Bengali agrarian imagery (fruit-bearing trees, life-giving rivers) will render vividly once the mandatory teaching note is attached.
• The “for my name’s sake” refrain gives a clean, repeatable teaching point against both karma-merit and weighed-deeds misreadings of grace, reinforcing the Romans baseline’s grace doctrine from an entirely different angle.

Recommended actions

1. Load assets/translation_memory.json (Ezekiel edition) and analysis/12_ai_translation_requirements.md (Ezekiel edition) before any Phase 2 segment work begins; both extend, and must never override, the Romans baseline files.
2. Run automated string-match QA on the three highest-frequency fixed phrases (Lord GOD compound, messenger formula, recognition formula) across every translated file before human review begins.
3. Route all 7 Critical-tier and 10 High-tier doctrines to mandatory human theologian review per doctrine_risk_registry.json; route the 3 Medium-tier doctrines (Holiness and Gradated Priestly Access, Inclusion of the Sojourner, Sabbath as Covenant Sign) to native-speaker review.
4. Attach the mandatory translator/teaching notes documented in 08_core_glossary.md and 11_doctrine_analysis.md at every flagged occurrence — especially the idolatry, nephesh, ruach-wordplay, “son of man,” and “my servant David” notes — before marking any segment complete.
5. Treat Ezekiel 37:1-14 as requiring theologian review in its entirety, regardless of individual term risk, given its status as this curriculum’s core passage.