Machine translation

Machine translation is the process by which a computer system produces, from a source-language computer-readable text, a target-language computer-readable text which is intended to be an approximate translation of the former.

Machine translation, a mature technology today, has radically changed the way in which people perceive multilingual communications, as nowadays anyone having access to the Internet can use it, for instance, to make sense of web content written in a different language. Of course, it has also impacted translation as a profession (and the way it is perceived by the general public). After defining machine translation and distinguishing it clearly from other computer-aided translation technologies and giving a brief historical review, from the early rule-based systems of the 1950s to the statistical systems of the 1990s and the early 2000’s to the advent of the “deep-learned” neural approaches in the twenty teens, this article describes how machine translation is used by ordinary people and in professional computer-aided translation workflows, and how it can be evaluated, both when considering adoption or during its development. It also describes the main technological approaches: on the one hand, rule-based machine translation and, on the other hand, corpus-based machine translation in its two flavours: statistical and neural, both to allow professional translators to make informed decisions about the technology and to raise the awareness of the general public about what to expect from this technology and how to use it where applicable. To close, some active research lines in the field of machine translation are outlined.