Classification without previous training, e. g. when using vector similarity to classify unseen documents.