THE PERFORMANCE OF NEURAL MACHINE TRANSLATION IN THE INDONESIAN TRANSLATION OF THREE FUNDAMENTAL CATHOLIC PRAYERS

Abstract

Literary translation is one of the greatest obstacles to neural machine translation development (NMT). NMT precision is susceptible to common issues in literary texts, such as lexical ambiguity, complex syntax, and structural grammatical constructions. This study investigates the literary translation of three Catholic prayers: "The Sign of the Cross," "The Lord's Prayer," and "Hail Mary." These objects have been selected for their distinctive linguistic characteristics, such as archaic vocabulary, uncommon structures, and unique line breaks. The purpose of this study is to evaluate the NMT's ability to overcome obstacles in literary translation based on the number of errors produced, a discussion of the errors, and the relative difficulty of their correction, as stated by Hutchins and Somers. On the basis of Koponen's theory emphasizing semantic accuracy, the errors produced by MTs are divided into two main categories: individual concept errors and relation between concepts errors. Subcategories are subsequently created from the two categories. The quantitative data indicates that the most common individual concept error is mistranslation, while the most common relation between concepts error is misunderstanding. The primary objective of this study is to evaluate the performance of NMT in translating the three Catholic fundamental prayers. The library and survey methods are used for this research. In library method, researchers compare multiple theories and related studies. In the meantime, for the survey, the researchers distribute questionnaires to respondents to assess the accuracy and readability of the NMT's translation.