Automatic translation of historical content as a method of counteracting social exclusion on the example of a selected educational application

Abstract

The subject of the article is the automatic translation of historical content as a method of counteracting broadly defined social exclusion, including the underappreciation of the histories of certain regions in the literature. This issue is illustrated using the example of translating a fifty-item sample of educational materials contained in a memorization-optimizing application based on the SuperMemo algorithm. For this purpose, the Generative Pre-trained Transformer (GPT) was used, specifically the gpt-4-0125-preview model. It was trained on data up to December 2023, and its context window is 128,000 tokens. The resulting translation turned out to be far from perfect, primarily due to errors in translating names and personal names (e.g., “Republic” as “Rzeczpospolita”), linguistic awkwardness (e.g., “It was then that the first unification of China occurred.”), and inconsistency in the use of articles (a/the). Despite this, the obtained result met the basic requirements set by the sample application, created according to the MVP concept, aiming to, with relatively low expenditure, practically check whether there is demand for a given product or service adequate to the costs of a full-fledged solution. The conducted study also contributes to considerations on the broader application of automatic translation, especially in areas even potentially affected by social exclusion.