English-Maithili Machine Translation and Divergence

Maithili is a resource poor language from India and Nepal. It lacks the basic computer tools necessary for its survival and promotion in the age of intense competition on the internet. Since English happens to be the dominant language in terms of content, and more than 90 percent of Indian speakers do not use English, a translation from the latter to the former is essential. This paper presents an initial attempt at creating a statistical English-Maithili Machine Translation (EMMT) system and a study of Machine Translation (MT) divergences and errors caused due to divergence.