Computational linguistics

A fixed word-level vocabulary lacks the name “Velkora.” How can a subword-tokenized model commonly represent it?