Towards a Participative Platform for Cultural Texts Translators

From Machine Translation to Human Translation

Machine translation (see Figure1) embeds a language theory in which translating could be reduced to applying a set of rules from a source language to a target language. First, this would require that a form could be replaced by another while preserving the meaning. On the contrary, both language theory and practice show that form-to-form translation (e.g. ‘London’ to ‘Londres’) is rare at any level (term to sentence). For example, when a translator cannot express a connotation in a translated form, she can move it on a neighbour form. Second, this would require the existence of language rules. For current theories, rules exist for genre but not for a whole language. In other words, they are not universal rules, but practical norms in use (Rastier, 2007). This last objection to rule-based machine translation could explain the renewed interest in computer-aided translation based on human translation. Among them, statistical machine translation is still not accurate enough to be used by professional, but, on the contrary, translation memories are widely used in translation agencies.

Figure 1.

Machine translation of a poem (screenshot)