Um estudo historiográfico da evolução da tradução automática

Abstract

Machine translation (MT) is a type of translation, which serves as a resource to facilitate the narrowing of language barriers. The emergence of first experiments with translation machines was in the 1940s in the USA, which had invested on studies in such area for almost two decades, until be impacted by a cut on investments in the late 1960s. The research resumed only in the 1980s and opened space for a new standard on scientific studies in the area. In the following decade, the opening of technical translation market promoted by offering of translation automation tools drove the progress of researches. Since the turn of the millennium, the studies on MT have demonstrated significant advances due to technological progress, which enabled the development of MT tools based on deep learning and artificial intelligence (AI). For reconstruction of such historiographical narrative, this work based on the concept of research programs proposed by Swiggers (1981) and theoretical specialty groups from Murray (1994), in order to establish the MT's research programs. As of an epistemological approach on the nature of translation from the language operations point of view (OUSTINOFF, 2003), which includes the cultural nature of translation process, we arrive at foundation of language mechanization methods (AUROUX, 1998). In order to build a historiographical interpretation of MT processes to establish a critical thinking about them, it was defined analysis criteria based on internal parameters (language incidence in translation adopted by MT models) and external parameters (historical and social context where linguistic ideas that influenced MT programs have moved around). The critical analysis seeks to highlight the theoretical and methodological disruptions that occurred on each MT's research program, as well as look at their changes from language view approach that determined the progress of MT models.