Tradução automática no Brasil: (des)automatizando números e discursos dos últimos 30 anos

Abstract

The first research in machine translation was inaugurated in the late 1940s, with the theory of linguistic code breaking by the mathematician Waren Weaver, and from this possibility, the topic has been investigated to the present day. However, these research movements were not continuous and suffered discouragement, especially with the ALPAC (Automatic Language Processing Advisory Committee) report, published in 1966, which concluded that machine translation was imprecise, expensive, and moved at a slow pace. The report slowed down research for almost a decade, until the topic was investigated again in the mid-1970s, with research divided into two main axes: theoretical studies regarding linguistics in the context of machine translation systems and studies dealing with computational tests, with empirical approaches, aimed at building these systems (HUTCHINS, 2007, 2015; STUPIELLO, 2010). With the technological evolution and popularization of the Internet, this subarea of Computational Linguistics has gained worldwide notoriety and has become a relevant research point for Translation Studies. Considering the importance of the scientific production on machine translation for Translation Studies (GILE, 2015), this paper investigated, through scientometric indicators (VANTI, 2002; ESQUEDA, 2020), combined with a qualitative bias, the scientific production on machine translation in the Brazilian scenario. Thus, it demonstrates which discussions have been held in the country over the past 30 years - i.e., from 1990 to 2020, and the impact of such discussions on the training of Brazilian translators. We collected all articles from Brazilian Translation Studies journals that dealt with the topic: “machine translation”. Thus, this search in the titles, abstracts or keywords of the publications resulted in 17 articles throughout all the editions of 14 verified journals. Due to the reduced number of publications, the search scenario was expanded and the same keywords were submitted to the Google Academic search platform, composing a research corpus with 87 publications. In addition to the quantitative research, which provided numerical data that determined the authors who have most published on machine translation in Brazil, their respective coauthors, the institutions to which these authors belong to, and the most recurrent correlated terms in the abstracts of these publications, this dissertation also took into consideration the content dealt with in this corpus with regard to machine translation from the Brazilian perspective.