Executive Summary: Luke Language Package for Kannada

Why it matters

Luke’s Gospel — anchored in Jesus’ Nazareth manifesto (4:16-21) — is the New Testament book most saturated with exactly the material Karnataka’s religious landscape makes hardest to translate safely: possession/exorcism narratives, ritual anointing language, an afterlife parable, a poverty-and-reversal ethic, and a caste-adjacent table-fellowship scandal repeated across five separate scenes. This package extends the proven Romans Language Package rather than starting over, but Luke’s narrative-parabolic form surfaces 62 new theological terms and 8 new curriculum doctrines that Romans, an epistle, never had to name. Getting the Nazareth sermon and its three echo passages (7:22; 19:9-10; 24:47) fixed and consistent is the single highest-leverage task in this whole package, since Luke returns to that passage’s vocabulary as its own thesis statement at three later narrative turning points.

Key findings

• Doctrine load: 36 total doctrines identified for Luke; 14 Critical and 18 High risk (32 of 36 require mandatory human theologian review — 89% of all doctrine content), versus only 2 Medium (native speaker review) and 2 Low (automated review).
• Term load: Of roughly 109 enforced terms in the updated translation memory (47 inherited verbatim from Romans, 62 new to Luke), approximately 22 are Critical-risk and 39 are High-risk — over half the entire glossary requires theologian-level scrutiny at every occurrence.
• Three genuinely new Critical collision zones beyond anything Romans required: (1) ಅಭಿಷೇಕಿಸು “anointed” (4:18) vs. Śaiva ritual liṅga-abhiṣeka; (2) ಪರದೈಸು/ನರಕ “Paradise/Hades” (16:19-31; 23:43) vs. temporary karma-purification cosmology (svarga/naraka); (3) ನಿಮ್ಮ ಮಧ್ಯದಲ್ಲಿ “kingdom within/among you” (17:21) vs. Lingayat iṣṭaliṅga-indwelling and Vedantic antaryāmin doctrine.
• Two Luke-specific syncretism risks absent from Romans: coastal Karnataka’s Bhūta Kōḷa spirit-propitiation tradition (relevant to every exorcism narrative) and caste-conditioned commensality norms (relevant to five separate table-fellowship scenes: 5:29-32; 7:36-50; 14:12-24; 15:1-2,23-32; 19:5-7).
• One irrecoverable lexical gap: the Greek ἄφεσις wordplay linking “release” (4:18) to “forgiveness of sins” (1:77; 3:3; 24:47) cannot be preserved in Kannada and requires mandatory translator notes at all four occurrences rather than a lexical fix.
• Existing Kannada Bible tradition is a genuine asset: all of Luke’s highest-risk terms build on already-established, non-ambiguous Romans-baseline renderings (ರಕ್ಷಣೆ, ಪುನರುತ್ಥಾನ, ದೇಹಧಾರಣೆ, ಕರ್ತ), so this package extends rather than reinvents the destination-language authority.

Risks

• Son of Man dilution risk: ಮನುಷ್ಯಕುಮಾರ occurs 25+ times — the single highest-frequency Critical term in the book — creating cumulative risk that time-pressured translation will flatten it toward a generic “a person” gloss somewhere in a long document, quietly erasing its Danielic titular claim.
• Joy-term drift: ಆನಂದ is the more “elevated” literary Kannada word for joy and will tempt stylistically-minded translators away from the mandated ಸಂತೋಸ, silently importing Vedantic/bhakti mystical-absorption connotations into Luke’s relational, historically-grounded rejoicing.
• Invisible-to-QA pragmatic risk: table-fellowship passages contain no lexically “wrong” words to catch — the risk is a flattened, generically hospitable translation that is lexically correct but theologically mute about the social scandal Jesus deliberately provokes.
• Core-passage sequencing risk: because 7:22, 19:9-10, and 24:47 all directly echo 4:16-21’s vocabulary, any Phase 2 workflow that does not finalize the core passage FIRST risks producing inconsistent, uncoordinated renderings across four separate high-visibility passages.

Opportunities

• Genuine points of positive contact, carefully bounded: Luke’s universal-scope doctrine (Jew/Gentile, rich/poor) resonates with the historic Lingayat movement’s own 900-year-old rejection of caste hierarchy, and Luke’s economic-reversal ethic (Magnificat, Sermon on the Plain, rich man and Lazarus) offers real cultural traction — provided teaching materials explicitly distinguish social resonance from doctrinal equivalence, per the comparative theology analysis.
• Reuse over reinvention: because 47 of the ~109 total enforced terms are copied verbatim from the proven Romans package, Phase 2 tooling and reviewer training can be substantially reused rather than rebuilt from scratch.
• Front-loaded risk clarity: this Phase 1 package has already pre-resolved the single most consequential ambiguity in the book (Luke 17:21) with a theologian-ready fixed rendering, removing what would otherwise be a mid-project blocking decision.

Recommended actions

1. Sequence Phase 2 so that Luke 4:16-21 and its three echo passages (7:18-23; 19:9-10; 24:47) are translated and theologian-approved first, before any other segment in the book, per the mandatory core-passage-first rule in 12_ai_translation_requirements.md.
2. Route all 32 theologian-review doctrines and both native-speaker-review doctrines to their assigned reviewers per doctrine_risk_registry.json before Phase 2 Step 17 sign-off; do not allow Critical/High content to reach automated-only review under time pressure.
3. Enforce the two irreducible-risk terms (ಅಭಿಷೇಕಿಸು “anointed” and ನರಕ/ಪಾತಾಳ “Hades”) with mandatory, non-skippable translator notes at every occurrence, since no safe lexical substitute exists for either.
4. Brief native-speaker reviewers explicitly on the ಆನಂದ/ಸಂತೋಷ joy-term distinction and the table-fellowship “social scandal” preservation requirement, since both are QA-invisible failure modes that lexical validation alone cannot catch.
5. Carry forward the Romans baseline’s forbidden-substitution list unchanged and add the eleven new Luke-specific forbidden substitutions (ಪ್ರಾಯಶ್ಚಿತ್ತ, ಆನಂದ-as-primary, bare ದೀಕ್ಷೆ, unqualified ಸ್ವರ್ಗ, unqualified ನರಕ, ನಮಸ್ಕರಿಸು, ನಿಮ್ಮ ಒಳಗೆ at 17:21, ವಚನ-for-Scripture, bare ಜ್ಞಾನ, literal ದ್ವೇಷಿಸು, ಶುಭ ವರ್ಷ) into the Phase 2 automated validation pipeline as machine-checkable rules.