Executive Summary — Ezra (English → Bengali)

Why it matters

Ezra opens the TRI Bengali library’s first Old Testament, Hebrew/Aramaic historical-narrative curriculum, built on top of the Romans (NT epistle) baseline. It carries a doctrinal risk profile the baseline never had to face: Bengal’s living Hindu ritual vocabulary supplies fluent, familiar words for nearly every concrete cultic noun in the book — altar, sacrifice, temple, priest, fasting — and Bengali Muslim theology’s fatalism vocabulary (তকদির) and Islamic covenant/scripture categories collide with Ezra’s providence and Scripture-authority claims. Getting Ezra 1:1-11 right sets the pattern for the sovereignty, worship-restoration, and separation-from-syncretism doctrines that run through all ten chapters.

Key findings

• 20 doctrines identified across all 10 chapters (full-book coverage confirmed, no chapter silently omitted): 5 Critical, 9 High, 5 Medium, 1 Low — matching doctrine_risk_registry.json (Ezra) exactly. 14 doctrines (Critical + High) require mandatory human theologian review.
• ~72 terms now tracked in the Ezra translation memory (13 inherited exactly from the Romans baseline + 46 new Ezra-specific terms + 13 proper names). Of the newly-registered term entries in assets/bible_term_registry.json, 3 are Critical and 25 are High risk — 28 terms requiring theologian review before Phase 2 lock.
• The single highest-consistency-critical new rule in the library is the ঈশ্বর/দেবতা fence (true God vs. pagan gods, Ezra 1:7) — more absolute than any single baseline rule because a single lapse inverts the book’s central sovereignty claim.
• The single highest syncretism-collision term identified across either curriculum to date is হোমবলি (burnt offering), which compounds two independently loaded Vedic/Shakta ritual morphemes (হোম, বলি).
• A wholly new anthropological fence — মন (human disposition) vs. আত্মা (Holy Spirit) — must be introduced and enforced everywhere Cyrus’s or the returnees’ “spirit” is stirred (1:1, 1:5), or the text risks implying the Holy Spirit indwelt a pagan king.
• পবিত্র বংশ (“holy seed,” 9:2) is the sharpest pastoral-doctrinal risk in the book: unguarded, it reads as scriptural endorsement of caste-based (Hindu gotra) or biradari (Muslim lineage) marriage restriction rather than covenant fidelity.

Risks

• Critical: Collapsing দেবতা into ঈশ্বর anywhere in chapters 1, 4-6 erases the monotheistic contrast that anchors the whole book.
• Critical: Rendering human ruach as আত্মা (1:1, 1:5) creates a pneumatological error and an unintended Hindu ātman association simultaneously.
• Critical: যজ্ঞবেদি/হোমবলি without their mandatory standing notes (ch.3, 6, 8) actively teaches that the OT sacrificial system is equivalent to ongoing Hindu ritual sacrifice.
• High: A single lapse into পুরোহিত (Hindu Brahmin officiant) anywhere across all ten chapters imports caste hierarchy onto the Aaronic priesthood.
• High: স্বেচ্ছাদান sliding toward bare দান (Ezra 1, 2, 8) reopens the grace-vs-merit battle the Romans baseline already fought and won for অনুগ্রহ.
• High (pastoral): Ezra 10:3, 10:44’s “put away foreign wives” narrated without careful historical-theological framing risks being taught as a timeless, transferable divorce principle.

Opportunities

• Ezra’s administrative/documentary genre (decrees, archives, genealogies, currency records) carries genuinely low doctrinal risk and can move through automated/native-speaker review quickly, freeing theologian capacity for the 14 Critical/High doctrines.
• The সদাপ্রভু covenant-name entry, once fixed here, becomes reusable infrastructure for any future OT curriculum (Nehemiah, Chronicles, the Prophets) sharing this library.
• The মন/আত্মা and ঈশ্বর/দেবতা fencing rules, once validated in Phase 2, become permanent additions to the shared forbidden-substitution list for all future OT material.

Recommended actions

1. Lock assets/translation_memory.json and assets/bible_term_registry.json (Ezra) before any Phase 2 segment translation begins; confirm both load alongside, not instead of, the Romans baseline files.
2. Treat দেবতা/ঈশ্বর, মন/আত্মা, and যজ্ঞবেদি/হোমবলি-with-standing-note as auto-reject (not merely auto-flag) validation failures, per 12_ai_translation_requirements.md (Ezra).
3. Route all 14 Critical/High doctrines and their associated 28 Critical/High terms to human theologian review; route the 5 Medium doctrines to native-speaker review; route the 1 Low doctrine (historical/administrative documentation) to automated review only.
4. Brief theologian reviewers specifically on the পবিত্র বংশ and (বিদেশী স্ত্রীদের) ত্যাগ করা pastoral-framing risk before chapter 9-10 review begins.
5. Confirm all Ezra proper-name transliterations (কোরস, দারিয়াবস, অর্তক্ষস্ত, etc.) against the Bangladesh Bible Society / Bible Society of India edition before Phase 2 lock, per 06_linguistic_gap_analysis.md §8.