Linguistic Gap Analysis
Linguistic Gap Analysis: Zephaniah (English → Bengali)
Curriculum: Zephaniah
Core passage: Zephaniah 3:14–17
PRD Phase 1 Step: 8 — Linguistic Gap Analysis
Governing authority: translation_memory.json (baseline, Romans) — all Section A terms below are REUSED, not renegotiated. This document extends the baseline for full-book Zephaniah coverage and must be read alongside 08_core_glossary.md and bible_term_registry.json (Zephaniah additions), which it supplies the underlying rationale for.
Coverage: Full book, Zephaniah 1:1–3:20. Every chapter reviewed; chapters/sections contributing no new terms are noted explicitly in Section 5.
1. Purpose and Method
Zephaniah presents a linguistic profile very different from Romans: it is Hebrew poetic prophecy rather than Greek epistolary argument, it uses the covenant name יהוה on nearly every page, and its central rhetorical device — the reversal of the same Day from wrath to joy — depends on Bengali vocabulary choices remaining stable across that reversal. This analysis identifies:
- Where existing Bengali Christian vocabulary (baseline + general Bengali Bible tradition) is available and adequate, with named weaknesses to guard against.
- Where Bengali vocabulary is genuinely missing and must be newly coined/selected.
- Where Bengali’s existing religious vocabulary is crowded by the dominant Hindu and Muslim semantic neighborhoods and requires deliberate “fencing” (context-anchoring, translator notes, or lexical avoidance) to prevent syncretistic misreading.
- Where the choice is transliteration vs. paraphrase, and why.
- A ranked list of this book’s highest-risk ambiguities for Phase 2 review routing.
2. Doctrine Vocabulary Matrix
| Doctrine | Available Bengali terms | Weaknesses of available terms | Recommended strategy |
|---|---|---|---|
| The Day of the Lord as Judgment and Restoration | সদাপ্রভুর দিন (new, from YHWH+day); সেই দিনে (echo formula); ক্রোধ/প্রচণ্ড রোষ (wrath) | No existing single Bengali idiom captures “one Day, two faces” — Bengali readers default to treating “day of judgment” (বিচারের দিন, familiar from Islamic yaum al-qiyamah idiom বিচারের দিন / কিয়ামত) and “a joyful day” as unrelated events. Risk of readers importing the Islamic Judgment Day (কিয়ামতের দিন) framework, which is retributive-only and has no restoration-for-the-humble structure. | Fix সদাপ্রভুর দিন as one invariant phrase across all 7+ occurrences (1:7,8,9,10,14×2,18; 3:16 echo). Never substitute বিচারের দিন. Require a structural teaching note at first occurrence (1:7) flagging that the SAME phrase returns in ch. 3 with a reversed tone, so the single-Day argument is visible across the whole book, not just within the core passage. |
| God’s Judgment on Idolatry and Complacency | প্রতিমাপূজা (idol-worship); বাল, মিল্কম (named false gods); আত্মতৃপ্ত (complacency); আকাশের বাহিনী (host of heaven) | প্রতিমাপূজা is the exact everyday Bengali term for the dominant Bengali Hindu religious practice (দুর্গাপূজা, কালীপূজা, প্রতিমা বিসর্জন are annual, publicly visible, culturally beloved events in both West Bengal and, to a lesser extent, Bangladesh’s Hindu minority). There is no neutral synonym that avoids this overlap entirely. | Do NOT avoid the collision by softening into vague “wrong worship” language — that would blunt the doctrine. Instead, FENCE it: always pair প্রতিমাপূজা with the specific named referents (বাল, আকাশের বাহিনী, মিল্কম) so the condemnation reads as historically specific (named ANE deities) rather than as a live commentary on a Bengali reader’s Hindu neighbors or relatives. Teaching notes must state this explicitly at 1:4-6. |
| The Remnant of Israel | ইস্রায়েলের অবশিষ্ট অংশ; আমার প্রজাদের অবশিষ্ট অংশ; দীন ও নম্র লোক; নম্রতা | অবশিষ্ট অংশ (“remaining portion”) is a flat, low-affect Bengali phrase with no inherent positive theological charge — it reads as neutral “leftover,” and Bengali has no established doctrinal loanword equivalent to the Hebrew שְׁאֵרִית’s rich prophetic resonance. Additionally, the SAME root is used for the condemned “remnant of Baal” (1:4), creating a genuine internal collision, not just an external one. | Retain অবশিষ্ট অংশ (no better native option exists) but require the doctrine to be carried by CONTEXT and by the definitional pairing with দীন ও নম্র লোক (3:12) rather than by the noun alone. A translator/teaching note at both 1:4 and 2:9/3:13 is mandatory, explicitly naming the shared-root collision so instructors address it rather than letting learners silently absorb it. |
| God’s Joyful Love for His People | ভালোবাসা (love); আনন্দ করা/উল্লাস করা/জয়ধ্বনি করা (joy cluster); পরাক্রমী (mighty warrior) | Bengali has abundant devotional-joy vocabulary from the Vaishnava bhakti tradition (কীর্তন, প্রেমানন্দ, লীলা-বিষয়ক আনন্দ) that could make God’s joy in 3:17 sound like Krishna-bhakti ecstatic devotional imagery — a Bengali reader may unconsciously assimilate “God rejoicing over his people” into the Radha-Krishna divine-romance/lila frame. প্রেম is rejected for exactly this reason (see glossary). | Use only ভালোবাসা (never প্রেম) for God’s love, and vary the joy-verb cluster (আনন্দ করা / উল্লাস করা / জয়ধ্বনি করা / জয়গান করা) rather than settling on a single term, specifically avoiding কীর্তন-adjacent vocabulary or any phrase that would cue bhakti ecstatic-devotion imagery. God must remain grammatically and theologically the Subject who loves and rejoices over a covenant people he has redeemed from judgment — not a divine lover figure in a romance narrative. |
3. Missing Vocabulary vs. Crowded Semantic Neighborhoods
Zephaniah’s vocabulary problems fall into two structurally different categories. Distinguishing them determines the correct fix: a MISSING term needs coinage/compounding; a CROWDED term needs fencing (context-anchoring and translator notes), not replacement.
3.1 Genuinely Missing Vocabulary (no adequate Bengali term exists)
| Concept | Gap | Resolution |
|---|---|---|
| שְׁאֵרִית (technical remnant-doctrine noun) | No single Bengali noun carries the prophetic-technical sense of “surviving, divinely preserved portion” distinct from ordinary “leftover.” | Compound phrase ইস্রায়েলের অবশিষ্ত অংশ, always used as a fixed unit, functioning as a proper doctrinal term rather than a literal description — parallel to how ধার্মিক বলে গণ্য হওয়া in the baseline compounds where no single word suffices. |
| הַקֹּפְאִים עַל־שִׁמְרֵיהֶם (wine-lees complacency idiom) | Bengali has no viticultural idiom of this kind (wine-making is not a live cultural referent in Bengal); a literal calque would be opaque or comic. | Abandon the image entirely; render the underlying sense with আত্মতৃপ্ত/নিশ্চিন্ত (settled self-satisfaction, practical unbelief in God’s active involvement). Teaching notes may explain the original wine image for cultural/historical interest but must not anchor the translation to it. |
| שֶׁכֶם אֶחָד (“one shoulder,” idiom for cooperative labor) | Bengali has no porter’s-yoke idiom with this specific figurative extension into “united worship.” | Paraphrase as একচিত্তে/একসঙ্গে সেবা করা (serve with one mind/together), sense-for-sense. |
| אַל־יִרְפּוּ יָדָיִךְ (“let not your hands be slack,” idiom for discouragement) | Bengali “hands” idioms exist (হাত-পা ছেড়ে দেওয়া, “to give up”) but do not map cleanly onto this specific restoration-oracle formula. | Render তোমার হাত শিথিল না হোক while allowing “do not lose heart/courage” as the controlling sense; retain the hand-image only where it reads naturally in the sentence, not as a rigid calque. |
| גֵּאֵי גַּבְהֵךְ (removal of the proudly exultant from among the people, as a purification event) | No single Bengali term for “the specific sub-group of proud people who get removed as a purifying act,” as opposed to pride as a general character trait. | Descriptive phrase অহংকারী/গর্বিত লোকদের দূর করা (removal, not just naming pride) to preserve the EVENT (purification) rather than only the TRAIT. |
| תְּהִלָּה וָשֵׁם (shame publicly reversed into praise and renown) | Bengali has vocabulary for shame (লজ্জা) and honor/fame (সুনাম, প্রশংসা) but no single fixed idiom for the specific reversal-of-public-status event this phrase names. | Descriptive phrase লজ্জা প্রশংসা ও সুনামে পরিণত হওয়া (shame turned into praise and renown) — a constructed phrase, flagged as new coinage for consistency across future OT curricula. |
3.2 Crowded Semantic Neighborhoods (adequate Bengali term exists, but is contested territory)
| Bengali term | Doctrine use in Zephaniah | Competing neighborhood | Fencing strategy |
|---|---|---|---|
| প্রতিমাপূজা | Idolatry judgment (1:4-6) | Standard, positively-connotated Bengali Hindu term for image-worship (দুর্গাপূজা etc.) | Always co-occur with named referents (বাল, আকাশের বাহিনী, মিল্কম); never used as a freestanding description of “other religions” in teaching material; frame historically (7th-century-BC Judah), not comparatively. |
| পুরোহিত | Israelite priestly office (1:4; 3:4) | Standard Bengali word for a Hindu ritual priest / Brahmin officiant | Context must always specify “সদাপ্রভুর মন্দিরের পুরোহিত” or equivalent covenant-office framing on first use per passage; never left to stand alone without the covenant qualifier in teaching text. |
| ঈর্ষা (jealousy, of God) | God’s covenant jealousy (1:18; 3:8) | Ordinary Bengali sense = petty envy, a vice among equals | Mandatory translator note at first occurrence distinguishing covenantal ঈর্ষা (a positive attribute of exclusive covenant loyalty) from vice-sense envy; consider pairing with অন্তর্জ্বালা (burning zeal) as a softening/clarifying doublet in teaching prose, though not necessarily in the base scripture rendering. |
| সদাপ্রভু vs. প্রভু vs. ঈশ্বর | YHWH (proper name) vs. NT κύριος title vs. generic Elohim | The three-way Bengali distinction is not intuitive to average readers, who may treat all three as interchangeable “God” synonyms, especially readers whose primary Bible exposure has been NT-only (many curricula, including this baseline’s, are Romans-first) | This is THE single highest-priority fencing task of this whole package (see §5, Rank 1). Explicit teaching-note requirement at 1:1 and repeated at first occurrence per chapter: সদাপ্রভু = the personal covenant name (translated as “the LORD” in English tradition, small-caps convention); প্রভু = reserved for the NT confession of Christ’s Lordship; ঈশ্বর = the generic word for God/deity. Never interchange. |
| বীর (rejected) vs. পরাক্রমী (chosen) | Divine warrior imagery (3:17) | বীর evokes Bengali/Indian epic heroes (Rama, Arjuna, regional folk-hero and warrior-deity traditions) and could suggest a demigod-hero category | পরাক্রমী (mighty, powerful) is selected specifically to denote might/capability without hero-cult resonance; never substitute বীর even in casual paraphrase or teaching commentary. |
| ভালোবাসা vs. প্রেম | God’s covenantal love (3:17) | প্রেম is saturated with Bengali romantic/cinematic connotation and, in religious registers, with Vaishnava bhakti divine-romance imagery | ভালোবাসা fixed as the only rendering; প্রেম never used even in devotional paraphrase materials, per baseline precedent (John 3:16 tradition). |
| আশ্রয় নেওয়া (take refuge) vs. বিশ্বাস (believe/trust) | Remnant’s refuge-seeking posture (3:12) | বিশ্বাস is the baseline’s fixed, heavily loaded term for saving faith in Christ (Romans); if used here it would suggest Zephaniah’s OT refuge-seeking IS New Testament saving faith, collapsing distinct lexical roots (חסה vs. אמן) into one | Keep আশ্রয় নেওয়া as a lexically distinct term; teach the CONCEPTUAL continuity (both involve reliant dependence on God) explicitly in commentary, without merging the vocabulary itself. |
| রক্ষা করা/উদ্ধার করা (save, verb) vs. পরিত্রাণ (salvation, noun) | Concrete historical deliverance (1:18; 3:17,19) | পরিত্রাণ is baseline-fixed for the NT soteriological noun (Romans); Zephaniah’s usage is the verbal root of the same Hebrew word family but describes concrete national deliverance from enemies/judgment, not the full NT doctrine of salvation | Use only the verb forms রক্ষা করা/উদ্ধার করা in the base rendering; reserve পরিত্রাণ as a noun exclusively for NT contexts; a bridging teaching note on root continuity (not homonymy) is required so the connection is taught deliberately rather than either ignored or over-collapsed. |
| ব্যবস্থা (law) | Priestly violation of Torah (3:4) | Crowded neighbors: ধর্ম (Hindu dharma/generic “religion”) and শরিয়ত (Islamic legal-idiom) | Already fenced by baseline; reaffirm exactly as inherited — no new fencing needed beyond baseline’s existing rule. |
| জাতি/জাতিসমূহ (nations, goyim) vs. পরজাতি (baseline Gentiles) | Judgment/restoration of the nations (ch. 2; 3:9-10) | পরজাতি is baseline-fixed for the NT Jew/Gentile ecclesiological contrast; using it here would import that specific NT theological frame into an OT nations-under-judgment-then-included-in-worship narrative that has a different shape | Use জাতি/জাতিসমূহ as the primary OT-context term; reserve পরজাতি for cross-reference/teaching-note use only, explicitly marked as a later NT development of a related but distinct theological trajectory. |
4. Transliteration vs. Paraphrase Decisions
| Item | Decision | Rationale |
|---|---|---|
| יהוה (YHWH) | Established-equivalent, not transliteration: সদাপ্রভু | This is not a phonetic transliteration of the Hebrew but the settled Bengali Bible tradition’s rendering (parallel to English “the LORD”). A phonetic transliteration (e.g., যাহওয়েহ) would break with all existing Bengali Bible usage and read as foreign/academic. Rejected: যিহোवा (Jehovah’s Witness translation-tradition association). |
| בַּעַל (Baal) | Transliterate: বাল | Must remain visibly a foreign proper name, not a translated common noun (its root literally means “lord/master,” but translating the root would wrongly suggest a legitimate title rather than a named rival deity). |
| מַלְכָּם (Milcom) | Transliterate: মিল্কম | Same logic as Baal — preserving the specific syncretism charge (swearing by YHWH AND a named rival god) requires the proper name, not a translated gloss (“their king”), which would obscure the accusation. |
| Proper place-names (Nineveh, Assyria, Cush, Moab, Ammon, Philistia, Cherethites, Sodom and Gomorrah, Zion, Jerusalem, Judah) | Transliterate/established form | All already have stable, established Bengali Bible spellings (see Section C of 08_core_glossary.md); no paraphrase question arises. |
| הַקֹּפְאִים עַל־שִׁמְרֵיהֶם (wine-lees idiom) | Paraphrase, not transliteration or literal calque | See §3.1 — the viticultural image has no live Bengali cultural referent; a literal rendering (তলানির উপর জমাট বাঁধা মদ) would be unintelligible or bathetic in a solemn prophetic oracle. |
| שֶׁכֶם אֶחָד (“one shoulder”) | Paraphrase | See §3.1. |
| אַל־יִרְפּוּ יָדָיִךְ (“hands not slack”) | Partial retention + paraphrase | The hand-image is retained (তোমার হাত শিথিল না হোক) because Bengali does have a related idiomatic register around hands and effort, but the controlling sense (do not lose heart) governs interpretation; full paraphrase is used as backup if the retained image reads awkwardly in a given teaching context. |
| אֲנִי וְאֶפְסִי עו�ד (Nineveh’s self-deification claim) | Near-literal retention, not loose paraphrase | See registry: this must be rendered literally enough (“আমিই আছি, আমার কেউ নেই”) to preserve the recognizable intertextual echo/parody of divine self-declaration formulas (cf. Isaiah 45:5); a loose paraphrase (“Nineveh was very proud”) would destroy the intended shock of a city arrogating God’s own self-description to itself. |
| שׁוּב שְׁבוּת (“restore fortunes/turn back captivity”) | Fixed idiomatic paraphrase: বন্দীদশা ফিরিয়ে আনা | This is a recurring fixed OT formula (also in Job, Jeremiah, Amos); render consistently as an idiom rather than word-for-word, and flag for cross-curriculum harmonization when those books are produced. |
| תְּהִלָּה וָשֵׁם and גֵּאֵי גַּבְהֵךְ | Descriptive paraphrase (new coinage) | No fixed Bengali idiom exists; constructed descriptive phrases are used and flagged as first-instance coinages requiring consistency tracking (see §3.1). |
| אַבָּא-style intimate address terms | Not applicable in Zephaniah | No Aramaic/intimate-address transliteration question arises in this book (unlike Romans 8:15’s আব্বা); noted here only to confirm the category was checked and found not present. |
5. Full-Book Chapter Review (Coverage Confirmation)
| Chapter | New terms / doctrinal load contributed | Status |
|---|---|---|
| 1:1-6 | সদাপ্রভু (superscription); বাল, আকাশের বাহিনী, মিল্কম, প্রতিমা-পুরোহিত, পুরোহিত (idolatry judgment); বালের অবশিষ্ট অংশ (condemned-sense remnant root collision established) | Reviewed — heaviest idolatry-vocabulary chapter in the book. |
| 1:7-13 | সদাপ্রভুর দিন (first full occurrences); ক্রোধ (wrath); আত্মতৃপ্ত (complacency) | Reviewed — establishes the Day-of-the-LORD technical term and the “second face of judgment” (complacency, not just active idolatry). |
| 1:14-18 | ক্রোধ cluster intensification (প্রচণ্ড রোষ/উত্তপ্ত ক্রোধ); রক্ষা করা/উদ্ধার করা (negative sense: silver/gold cannot deliver); ঈর্ষা (jealousy) first occurrence; রৌপ্য ও স্বর্ণ | Reviewed — closes ch.1 on total, inescapable judgment; no restoration language yet, by design. |
| 2:1-3 | নির্লজ্জ জাতি (shameless nation); সদাপ্রভুর অন্বেষণ করা (seek the LORD, central positive imperative); দেশের নম্র লোকেরা (humble of the land); নম্রতা paired with ধার্মিকতা (reused baseline); আশ্রয় পাওয়া (shelter in the day of wrath) | Reviewed — pivot chapter: first positive imperative and first appearance of the humility/remnant seedbed vocabulary. |
| 2:4-15 | জাতি/জাতিসমূহ (nations under judgment: Philistia, Moab, Ammon, Cush, Assyria/Nineveh); সদোম ও ঘমোরা (typological reference); গর্ব/অহংকার (pride, Moab/Ammon and Nineveh); আমার প্রজাদের অবশিষ্ট অংশ (FIRST positive remnant statement, 2:9); আমিই আছি আমার কেউ নেই (Nineveh’s self-deification claim) | Reviewed — the international oracles chapter; establishes both the pride/humility antonym pair at national scale and the doctrinally pivotal first positive remnant phrase. |
| 3:1-5 | গর্জনকারী সিংহ/সন্ধ্যার নেকড়ে (corrupt-leadership predator imagery); ব্যবস্থার প্রতি সহিংসতা (violence to the Law, reusing baseline ব্যবস্থা); পবিত্রস্থান অপবিত্র করা (profane the sanctuary, reusing baseline পবিত্র); ধার্মিক (YHWH’s own righteous character, reusing baseline root) | Reviewed — fourfold leadership indictment; theological ground of hope (God’s unchanging righteous character amid human corruption) established here, feeding directly into ch.3’s later restoration turn. |
| 3:6-8 | ক্রোধ (wrath, echoed); নেশন-judgment recap; no new Section B terms beyond those already logged | Reviewed — transitional recap; confirmed no new load-bearing vocabulary beyond items already captured in 1:14-18 and 2:4-15. |
| 3:9-10 | বিশুদ্ধ ভাষা (pure speech, restored to the NATIONS); একচিত্তে সেবা করা (serve with one accord); আমার বিক্ষিপ্ত লোকদের কন্যা (daughter of my dispersed ones) | Reviewed — universal-scope restoration promise extending beyond Israel; important doctrinal counterweight to ch.2’s judgment-of-nations material. |
| 3:11-13 | অহংকারী/গর্বিত লোকদের দূর করা (removal of the proud, purification of the remnant); দীন ও নম্র লোক (definitional remnant statement); ইস্রায়েলের অবশিষ্ট অংশ (named doctrinal term, definitive occurrence); সদাপ্রভুর নামে আশ্রয় নেওয়া (take refuge in YHWH’s name) | Reviewed — the doctrinal core of the Remnant of Israel doctrine; every Critical-risk remnant term converges here. |
| 3:14-17 (CORE PASSAGE) | সিয়োন-কন্যা, সিয়োন, যিরূশালেম, ইস্রায়েলের রাজা (King of Israel/YHWH), পরাক্রমী (mighty warrior), ভয় করো না, তোমার হাত শিথিল না হোক, আনন্দ করা/উল্লাস করা/জয়ধ্বনি করা/জয়গান করা (joy cluster), ভালোবাসা, তিনি তাঁর ভালোবাসায় প্রশান্ত/নীরব থাকবেন (MT textual-variant term), সেই দিনে (echo of Day-of-the-LORD formula in restoration key) | Reviewed — theological anchor of the whole curriculum; every term here converges terms established across chs. 1-3 into a single reversal-of-the-Day scene. |
| 3:18-20 | বন্দীদশা/অবস্থা ফিরিয়ে আনা (restore fortunes); বিক্ষিপ্ত লোকদের সংগ্রহ করা (gather the outcast); লজ্জা প্রশংসা ও সুনামে পরিণত হওয়া (shame turned to praise/renown) | Reviewed — closing regathering and honor-reversal oracle; completes the restoration arc opened at 2:9 and 3:9-10. |
Coverage confirmation: All 53 verses of Zephaniah (1:1–3:20) have been reviewed. No chapter or section was silently omitted; 3:6-8 is explicitly noted as contributing no new load-bearing vocabulary beyond terms already logged elsewhere.
6. Ranked List of Highest-Risk Ambiguities
Ranked by combined severity (doctrinal centrality × likelihood of collapse into a competing religious framework) for Phase 2 review-routing priority.
-
সদাপ্রভু vs. প্রভু vs. ঈশ্বর three-way distinction. The single highest risk in this package. If Phase 2 translators or reviewers trained primarily on the Romans baseline (where প্রভু is heavily reinforced for Christ’s Lordship) default to প্রভু for יהוה, the OT proper-name/NT title distinction collapses across nearly every verse of the book. Mandatory human theologian review at first occurrence of every chapter.
-
সদাপ্রভুর দিন consistency across the judgment/restoration reversal (1:7 through 3:16). If rendered with even minor phrasing variation, or if the ch.1 and ch.3 occurrences drift apart lexically, the book’s single central argument — one Day, two faces — becomes invisible to a Bengali reader encountering the chapters separately in a teaching sequence.
-
প্রতিমাপূজা’s collision with living, publicly celebrated Bengali Hindu devotional practice. Highest social-sensitivity risk in the book, given that this term names the reader’s own neighbors’ or family members’ religious practice in West Bengal specifically. Requires deliberate historical fencing at every occurrence (1:4-6), not vocabulary substitution.
-
অবশিষ্ট অংশ shared root between condemned “remnant of Baal” (1:4) and the salvific “remnant of Israel/my people” doctrine (2:9; 3:13). An internal (not external-religion) collision — Bengali offers no lexical way to distinguish the two without context, making this the highest INTERNAL-consistency risk in the glossary.
-
ভালোবাসা / joy-cluster vocabulary drifting toward Vaishnava bhakti devotional-romance registers in 3:17. The doctrinal centerpiece of “God’s Joyful Love for His People” is exceptionally vulnerable to being heard as Krishna-lila divine romance imagery if প্রেম or কীর্তন-adjacent vocabulary creeps in during oral teaching or worship-song adaptation, even if the base scriptural text avoids it correctly.
-
King of Israel title (3:15) inviting a premature Christological identification. Teaching materials must resist collapsing this OT declaration of YHWH’s own kingship into a direct Christological statement, while still permitting the doctrinal trajectory to be taught — a fine line requiring theologian oversight.
-
ঈর্ষা (jealousy of God) read as petty envy/vice. Without the mandatory translator note, this is the term most likely to produce an actively negative impression of God’s character among readers unfamiliar with covenant-jealousy theology.
-
রক্ষা করা/উদ্ধার করা (save, verb) vs. পরিত্রাণ (salvation, noun) root-continuity teaching. Risk runs in both directions: under-teaching the connection makes Zephaniah’s deliverance look unrelated to NT salvation; over-collapsing the vocabulary makes OT national deliverance sound identical to NT soteriology, eliding real differences in scope and referent.
-
আশ্রয় নেওয়া (take refuge) vs. বিশ্বাস (baseline faith/trust in Christ). Risk of instructors, in the interest of building a “faith” bridge to NT doctrine, informally substituting বিশ্বাস for আশ্রয় নেওয়া in oral teaching even where the base text correctly avoids it.
-
পুরোহিত (priest) read through the lens of a Hindu Brahmin/ritual-priest role, especially in 3:4’s indictment of corrupt priests — risk that the doctrinal weight (violation of a divinely instituted covenant office) is read instead as generic anti-clerical sentiment recognizable from Bengali social commentary on ritual specialists.
-
আকাশের বাহিনী (host of heaven) intersecting with live Bengali folk-astrological practice (জ্যোতিষ, রাশিফল). Lower doctrinal centrality than items above, but real risk of the condemnation being read as a blanket statement against astronomy/astrology broadly rather than the specific ancient astral cult in view.
-
বীর/পরাক্রমী discipline in 3:17’s divine-warrior imagery. Lower risk than items above because a clear rejected-alternative has already been fixed in the glossary, but requires ongoing vigilance in any future hymn, liturgy, or teaching-illustration adaptation where বীর may creep back in for poetic effect.
7. Summary Recommendation for Phase 2
This book’s linguistic profile concentrates risk far more heavily in proper-name/title discipline (সদাপ্রভু/প্রভু/ঈশ্বর) and cross-verse consistency (সদাপ্রভুর দিন; অবশিষ্ট অংশ) than in single-word mistranslation risk of the kind the Romans baseline was chiefly built to guard against. Phase 2 tooling must therefore add a cross-occurrence consistency validator (verifying identical phrasing for সদাপ্রভুর দিন and ইস্রায়েলের অবশিষ্ট অংশ at every occurrence) in addition to the term-level forbidden-substitution checks already specified in 12_ai_translation_requirements.md. All Critical and High items identified above must be loaded into an updated translation_memory.json and bible_term_registry.json before any Zephaniah segment enters Phase 2 translation.