Langages et savoirs : intelligence artificielle et traduction automatique dans la communication scientifique

The relationship between translation and AI is analysed from two different perspectives, scientific and linguistic, in order to provide a frame of reference on these issues as well as suggestions on the connections between languages and knowledge in AI perspective. After an analysis of the role of language in scientific communication, and of the overcoming of the use of a international language to the different national languages thanks to the diffusion of automatic translation, the attention is focused on the importance of terminology in scientific communication and on the interest of its correct use within digital processing. Case analyses are illustrated, applied to articles translated into different languages on the websites of a number of periodicals, which have undergone human review. The analyses carried out help to detect the main differences between machine translation and translation with human intervention: from a quantitative point of view, phenomena at the lexical level are the most important source of errors. In the case of specialised texts, the need for thorough revision arises, particularly in the area of terminology. In some cases, the incorrect or inconsistent use of terms significantly alters the message content of the original text.