Protected terminology
Mark mandatory, forbidden, deprecated, and context-limited translations.
Feature
Guardrails
Define what must be preserved, avoided, reviewed, or escalated in specialized translation work.
Quality rules for sensitive language.
Some terms must never be translated literally. Others require jurisdictional notes, approved equivalents, controlled register, or domain expert review.
Guardrails combine glossary status, ontology relationships, document type, target audience, and compliance needs into practical checks for translators and reviewers.
Review triggers
Escalate legal, medical, religious, safety-sensitive, or reputationally sensitive terms automatically.
Audience constraints
Define when a term may be expert-only, public-facing, simplified, or learner-friendly.
Policy memory
Turn previous decisions into repeatable checks instead of relying on individual recall.
What this adds
Rules become visible before publication.
Guardrails help teams catch problems while text is still being drafted, reviewed, or adapted.
Terminology safety
Flag forbidden equivalents, missing preferred terms, and outdated memory matches.
Domain sensitivity
Require review for terms tied to law, medicine, religion, safety, identity, or contested history.
Register control
Prevent expert-only language from appearing in plain-language notices without explanation.
Connected records
Guardrails draw from the whole knowledge layer.
The best checks are not generic grammar rules. They reflect the customer's domain, audience, and approved history.
Glossary status
Approved, rejected, deprecated, conflict, and needs-review states drive checks.
Ontology constraints
Jurisdiction, period, institution, concept type, and relationship rules shape recommendations.
Workflow stage
Different rules can apply to draft, expert review, final approval, and publication.
Source authority
High-authority sources can require stricter preservation than informal examples.