Low Resource Translation for Middle Egyptian

In his previous work, the author applied recent neural machine translation methods to Middle Egyptian, a 4000 years old language and one of the first languages overall. The application of neural translation techniques on this special resource is a rare combination and probably the first work in the field. The current aim is to apply Multi Task Learning (MTL) together with transliteration/POStags and incorporate CNN encoding to enhance translation quality generated solely by hieroglyphs. In future, back translation with the encoder trained during MTL will be deployed to multiply the current available hieroglyphic data.