Numbers → Bengali Language Package: Executive Summary

Why it matters

Numbers is the first Old Testament curriculum built on this Bengali Language Package, and it introduces the package’s first sustained encounter with the Hebrew Masoretic text, the OT covenant-name convention (সদাপ্রভু), and — in the bronze serpent of Numbers 21:4-9 — the single highest-collision translation risk yet documented in this pipeline: Bengal’s living Manasha (মনসা) snake-goddess healing tradition sits directly on top of the core passage’s own subject matter (snakebite, a serpent image, healing). Getting this passage right is the pipeline’s proof case for handling OT-to-Christ typology in a Bengali cultural context; getting it wrong risks converting the curriculum’s central teaching image into an accidental endorsement of the very folk practice it must be distinguished from.

Key findings

• Full-book coverage complete. All 36 chapters of Numbers have been mapped to doctrine, term, and risk (analysis/11_doctrine_analysis.md); no chapter was silently omitted — even administrative/census chapters (2, 7, 34) are logged as reviewed with no new doctrinal content.
• 51 new or elevated-risk terms were identified beyond the Romans baseline and are now formally seeded into assets/translation_memory.json, alongside 17 baseline terms carried forward unchanged (“inherited from Romans package”).
• 23 named doctrines span the seven curriculum-specified doctrines plus 16 supporting doctrines required for full-book coverage (assets/doctrine_risk_registry.json).
• The single highest risk item in the entire package is not a doctrine but a specific verse pair: Numbers 21:8-9’s দেখা (“look”) and বাঁচবে (“live”) must never be rendered as বিশ্বাস করা (“believe”) or পরিত্রাণ পাবে (“receive salvation”) in the base narrative text — a self-inflicted-error risk ranked above even the Manasha collision itself in analysis/06_linguistic_gap_analysis.md §6.
• The covenant name সদাপ্রভু is a necessary, non-contradicting extension of the Romans baseline’s প্রভু entry, required throughout Numbers and every subsequent OT book this pipeline will process.

Risks

Risk tier	Doctrines	New/inherited terms requiring theologian oversight
Critical	3 (Unbelief and the Wilderness Generation; The Bronze Serpent as a Type of Christ; Atonement and Priestly Mediation)	6 new (সদাপ্রভু, পিতলের সাপ, অবিশ্বাস, প্রায়শ্চিত্ত, প্রায়শ্চিত্তের দিন) + 4 inherited (ঈশ্বর, পরিত্রাণ, প্রভু, মেসায়া-class typology terms)
High	14	21 new + 8 inherited (elevated in this curriculum: মধ্যস্থতা, ভাববাদী, ভাববাণী)
Medium	6	15 new + 2 inherited (native speaker review)
Low	2	1 new (automated review only)
Total requiring human theologian review	17 doctrines / 39 terms (Critical + High combined)	Every occurrence, no sampling
Total requiring native speaker review	6 doctrines / 15 terms	—
Total automated-only	2 doctrines / 1 term	—

The three highest-severity, non-hypothetical collision risks (ranked in analysis/06_linguistic_gap_analysis.md §6):

1. পিতলের সাপ vs. Manasha (মনসা) snake-goddess healing folk practice — situational, not merely lexical, collision.
2. দেখা/বাঁচবে vs. বিশ্বাস করা/পরিত্রাণ পাবে — risk of translator over-correction collapsing OT type into NT fulfillment inside the base text itself.
3. প্রায়শ্চিত্ত vs. Hindu karmic self-penance — recurs at high frequency (ch. 16, 25, 28-29) and touches the same territory as Grace/Justification, already Critical in the Romans baseline.

Opportunities

• Deliberate lexical irony, ch. 1 vs. ch. 22-24: গণনা (God’s true census of his people) and ভাগ্যগণনা (Balaam’s false divination) share a root; this can be taught as an intentional literary-theological contrast rather than merely fenced defensively.
• Cross-curriculum vocabulary continuity: অবিশ্বাস is built directly on the Romans baseline’s Critical term বিশ্বাস, letting learners carry a single faith/unbelief vocabulary across both Testaments studied in this pipeline.
• Established-term precedent reuse: সদাপ্রভু, প্রায়শ্চিত্ত, হোমবলি, and নৈবেদ্য all follow the same historically-rooted-but-doctrinally-stabilized pattern the Romans baseline already established for ঈশ্বর — reducing the need for wholesale new-term invention despite Numbers’ extensive cultic vocabulary.
• Reusable OT extension pattern: the সদাপ্রভু/প্রভু distinction, the Hebrew-primary-original methodology, and the “commentary-layer-only” discipline for typological terms (grace, salvation, look/believe) are all designed for direct reuse by future OT curricula in this pipeline, not one-off Numbers solutions.

Recommended actions

1. Adopt assets/translation_memory.json (Numbers extension) and analysis/12_ai_translation_requirements.md as mandatory Phase 2 pre-flight load, jointly with the Romans-baseline artifacts — not as a replacement.
2. Hard-code the two highest-ranked validation checks (bronze-serpent translator note; দেখা/বাঁচবে substitution ban) into the Phase 2 AI system prompt as non-bypassable checks, per analysis/06_linguistic_gap_analysis.md §7 recommendation #2.
3. Route all Critical and High-risk Numbers segments to human theologian review at 100% occurrence rate, not sampling — this matches the Romans baseline’s own routing standard and is non-negotiable given the doctrinal centrality of the core passage.
4. Brief theologian reviewers specifically on the Manasha/tabij (folk-amulet) collision and the karmic-atonement collision before Phase 2 begins, since both require cultural literacy beyond standard biblical-Hebrew training to catch confidently.
5. Carry the সদাপ্রভু/প্রভু distinction, Hebrew-primary methodology, and commentary-layer discipline forward as reusable house-style rules for every future Old Testament curriculum processed in this Bengali Language Package.

---

Full supporting detail: analysis/07_semantic_analysis.md (core passage + full-book term study), analysis/08_core_glossary.md (glossary), analysis/04_comparative_theology.md (comparative religion detail), analysis/06_linguistic_gap_analysis.md (gap analysis and risk ranking), analysis/11_doctrine_analysis.md (full-book doctrine mapping), assets/translation_memory.json, assets/bible_term_registry.json, assets/doctrine_risk_registry.json, analysis/12_ai_translation_requirements.md.