Core Glossary

translation_memory.json is the enforced glossary for every Phase 2 translation in this curriculum. This document summarizes its shape and the principles behind it; see the Glossary Risk Groups for the full per-term entries.

Composition

The glossary currently holds 47 terms spanning all four risk tiers, drawn from the doctrines identified in Doctrine Analysis and grounded in the philosophical risks identified in Culture Analysis and Comparative Theology. Every term entry records:

• The approved Sanskrit translation and transliteration
• The doctrine risk tier (Critical/High/Medium/Low)
• Explicitly rejected alternatives, with the specific classical school or text each rejected term belongs to
• Notes explaining the non-obvious translation choice, citing primary sources where applicable (Bhagavad Gita, Upanishads, Mimamsa/Nyaya technical vocabulary)

Governing principles

1. No neutral word exists — select the least doctrinally contradictory available technical term. Unlike vernacular Language Packages, which can often find or coin a comparatively unloaded word, nearly every Sanskrit candidate is already a precise technical term in a developed system. The governing discipline is deliberate selection among imperfect options, not the search for a clean one.
2. Mandatory redefinition on use for Critical/High terms. Because this audience is trained to actively parse philosophical vocabulary, silent contextual disambiguation is not sufficient; Critical and High risk terms require explicit redefinition at first occurrence in every document.
3. Explicit rejection with citation, not silent avoidance. Every Critical-risk term records which specific school or text the rejected alternative belongs to (see alternatives_rejected and notes), so a reviewer can verify the reasoning against the primary source rather than trusting a general cultural claim.
4. Reuse existing Christian Sanskrit precedent where it exists. Proper names follow the 1808 Serampore Sanskrit New Testament; the God-term follows Roberto de Nobili’s Sarvesvara precedent. Where no precedent exists, this glossary makes a reasoned, documented, precedent-setting choice rather than assuming settled usage.
5. Version-controlled and append-only in Phase 2 — if a new term is discovered during document translation, it is added to translation memory and the version number incremented, never silently improvised per-document (see the AI Translation Requirements’ Translation Memory Load and Enforcement Instructions).

Relationship to the Doctrine Risk Registry

Every glossary term’s doctrine field links back to an entry in doctrine_risk_registry.json, so a term’s risk tier is always traceable to the specific doctrine it protects, and from there to the specific classical text or school the doctrine risk registry cites as the source of the collision.