Microsoft Translator Hub for Māori language

Abstract

Recent improvements in Machine Translation (MT) software has opened new possibilities for applications of automatic language translation. But can these opportunities exist for the smaller, minority languages of the world? Training data was collected for the language pair of Māori and English which was used to build an MT system using Microsoft’s Translator Hub software. A comparative analysis was undertaken with this system and Google Translate. Various MT metrics and analysis software was considered before deciding to use the Asiya toolkit to undertake the comparative analysis. Māori language experts were also used to provide a human perspective on the output translations. The overall results showed no significant difference in the translation quality produced by the two systems. Despite some misgivings around the accuracy of the translations these results do show promise for MT usage by minority languages.