← Back to glossary
Browse letter Z hub
Zero-Shot Translation
Zero-shot translation uses shared multilingual representations to translate new language pairs without direct parallel supervision.
Definition
A multilingual translation capability where models translate between language pairs not directly present in training data.
How It Works
Zero-Shot Translation helps teams build predictable AI and translation workflows by setting clear expectations for quality, consistency, and decision-making.
In production environments, this concept is applied with process controls such as human review, terminology alignment, and repeatable quality checks across multilingual content.
Key Concepts
- core principle of zero-shot translation
- workflow-level implementation
- terminology and quality consistency
- human validation before publication
Where It Is Used
- localisation workflows
- AI translation pipelines
- multilingual content production
- cross-referencing related concepts such as Zero Data Retention