kbd-translator

Kabardian↔Russian bidirectional translation uses a corpus-first approach, prioritizing patterns from 292k verified translation pairs. It translates in both directions (Russian→Kabardian and Kabardian→Russian) while maintaining proper Kabardian SOV word order and morphology, falling back to dictionaries when corpus data is insufficient.

Corpus-first translation prioritizes real-world translation patterns extracted from verified language pairs over generalized rules. This approach ensures translations reflect actual usage patterns and contextual accuracy, with dictionary lookup only as a fallback when corpus coverage is incomplete.

Yes. The Skill enforces Kabardian SOV word order and morphological rules during generation, ensuring outputs follow grammatical constraints specific to Kabardian even when translating from Russian source text.

When the 292k verified translation pairs don't contain matching patterns, the Skill falls back to dictionary lookup. This hybrid approach ensures coverage for both common phrases and edge cases, though dictionary results may lack contextual nuance.

Corpus-derived translation learns from actual translation examples rather than applying predetermined grammar rules. It discovers how native speakers handle real-world phrases, producing more natural outputs and capturing idiomatic patterns that rule-based systems often miss.