Abstract
Many Nigerian languages have relinquished their previous prestige and purposein modern society to English and Nigerian Pidgin. For the millions of L1speakers of indigenous languages, there are inequalities that manifestthemselves as unequal access to information, communications, health care,security as well as attenuated participation in political and civic life. Tominimize exclusion and promote socio-linguistic and economic empowerment, thiswork explores the feasibility of Neural Machine Translation (NMT) for the Edoidlanguage family of Southern Nigeria. Using the new JW300 public dataset, wetrained and evaluated baseline translation models for four widely spokenlanguages in this group: \`Ed\'o, \'Es\'an, Urhobo and Isoko. Trained models,code and datasets have been open-sourced to advance future research efforts onEdoid language technology.