Abstract: This paper discusses the philosophical question on whether understanding in human style is necessary for machine translation. Although it remains unsolved what human understanding exactly means in science and in philosophy, this paper points out the success of state of the art machine translation methods roots in compositionality of human languages. As long as the machine can “learn” and “understand” the mapping and the compositionalities of two languages from bilingual training corpora, machine translation can achieve good quality. In other words, machines can have their own understanding of human languages, different from human understanding.

Key Words: Machine translation; Understanding; Ambiguity

This article can be downloaded here

https://jdn.ucas.ac.cn/public/uploads/files/67b6e1d816fe4.pdf