TNU converts words into numeric and ontological vectors to compute semantic similarity, offering deterministic and explainable translations. It uses a hybrid search combining BLAKE2b identifiers and ontology layers to rank candidate translations. Designed for researchers and developers needing transparent multilingual mapping, it outperforms black‑box translators by exposing relations like same, related, broader, and opposite. The Python implementation includes tests, batch reporting, and a SQLite ontology database.
View on GitHub →jessfrt/TNU