Automatic Translation: Background, Problems and Perspectives

The term ‘Machine Translation’ – commonly abbreviated MT – is historical and polysemous. Historically it points back to pre-computer times with occasional engineering attempts at developing mechanical translating devices (Hutchins 1986; 1995). An earlier synonym, ‘mechanical translation’, is not used any longer, but ‘machine translation’ remains in common use, even if ‘computer translation’ would be more specific, and ‘algorithmic translation’ more precise, since it is algorithms running on computers that control the translation process – or its equivalent, as we will see shortly. In opposition to this, with humans in control, in MT jargon we speak of ‘human translation’. We thus have a fundamental polarity: “man translates” vs. “machine, i. e. computer translates” – with, in practice, mixed-modes between these extremes. Machine translation with human participation – also dubbed ‘interactive’ when it takes place during the translation process proper – is known as ‘human-assisted machine translation’ (HAMT); and human translation with computer support, nowadays the rule, is known as ‘machine-assisted human translation’ (MAHT) or, more commonly and briefly, ‘computer-assisted translation’ (CAT).