Ilia-Iliev/romani-sketches-dpo

This dataset consists of parallel text pairs for translating sentences from colloquial spoken Romani to Bulgarian. The style is conversational and the samples are taken from short dialogue videos. It is formatted as instruction-completion pairs suitable for training machine translation models.