Reading Political Economy at the Level of the Sentence: Automated Translation and the Global South

Abstract

Automated translation is being touted as revolutionizing inter-lin­gual communication on a global scale. Models such as Google Translate and DeepL are claimed to show a high accuracy rate in translating between cer­tain languages, often to the extent of surpassing human-level performances. How­ever, automated translation between most of the world’s languages, pri­mar­ily those from the Global South, does not exhibit a consistently high stan­dard accuracy rate. The reason for this disparity is discussed only su­per­fi­cially by the creators of these models. It is either claimed to be the result of the limited online presence of these languages or is written off as a symptom of these languages’ inherent complexities. This article attempts to un­der­stand the underlying structures that constitute this inequality in the field of au­tomated translation. In light of world-systems analysis, this article argues that the underperformance of automated translation technology in certain lan­guages is a systematic project of subordination of the Global South. It de­lineates the chain of processes, starting with European colonialism, that led up to the modern-day invention of automated translation, to examine how this inequality was at the core of global events, although it has always been branded as organic. This article then shows how this new ‘milestone’ of automated translation also becomes a device that operates to propagate the ideologies of the dominant structures of power.