- Orevaoghene Ahia and Kelechi Ogueji, colleagues at InstaDeep in Nigeria, developed a pidgin-to-English machine translation project that was accepted at the NeurIPS conference.
- They faced challenges in finding data for the translation but trained an Unsupervised Neural Machine Translation model to create a pidgin-English word pair catalogue from scratch.
- Their work aims to make technology more inclusive for the 75 million Nigerians who speak pidgin English and could potentially help improve sentiment analysis in digital marketing.
No comments yet
Be the first to share your opinion!