An empirical study of the scope of Spanish to Chinese machine translation. A GNMT case study concerning metaphorical and metonymic expressions

Abstract

Machine translation is a subject of great interest. GNMT (Google Neural Machine Translation), which, according to its maker Google, uses deep learning, can reduce the error rate by 60%. But is GNMT a true “cure-all” for translation woes? How does it work in terms of translating between a pair of the most spoken world languages: Chinese and Spanish? Above all, how does it work with regard to the translation of metaphorical and metonymic expressions, since natural language is full of such kinds of expressions? In this article we will try to answer these questions. Through a small sampling of 37 metaphorical and metonymic expressions based on the noun “head”, we will observe the correlation between a phrase’s frequency of use and its corresponding quality of translation. Through the t-test, we were able to verify the hypothesis that the higher the frequency of use, the better the resulting.