THE PERFORMANCE OF GOOGLE TRANSLATE IN TRANSLATING THREE CATHOLIC FUNDAMENTAL PRAYERS

Authors

  • Marchelline Berliandika Saksono Universitas Sanata Dharma
  • Alma Anindita Universitas Sanata Dharma
  • Harris Hermansyah Setiajid Universitas Sanata Dharma

DOI:

https://doi.org/10.25170/kolita.20.3797

Keywords:

Catholic prayers, error analysis, Google Translate, Koponen error category, machine translation

Abstract

Literary translation is one of the most difficult areas of machine translation (MT) development. Accuracy in MT is susceptible to issues that are frequently encountered in literary works, such as lexical ambiguity, syntax complexity, and structural grammatical constructs. This study examines the literary translation of three Catholic prayers: the "Sign of the Cross," "The Lord's Prayer," and "Hail Mary." These objects were chosen for their language's unique characteristics, which include antiquated vocabulary, strange structures, and unusual line breaks. The research is conducted to determine the MT's ability, that is represented by Google Translate (GT),  to overcome hurdles in literary translation as measured by the number of errors made, their discussion, and their relative difficulty rectification based on Hutchins and Somers' assertions. The quality of a translation, whether human or computer, has long been a subject of discussion, with no universally accepted metric. There are a variety of factors to consider based on the circumstances around the exercise, including accuracy, naturalness, fluency, and function. For machine translation, which is frequently used to obtain quick information about a document, semantic accuracy should definitely take precedence over fluency as Koponen states.The research attemps to measure GT’s performance in translating three Catholic prayers using Koponen’s error category. Koponen's notion of semantic accuracy categorizes the errors produced by the MT into two broad categories: individual concept errors and relational concept errors. The two groups are further subdivided. The results indicate that GT is underperformed when confronted with the individual concept errors, but performs admirably when the ST line is straightforward and well-structured. GT made a total of 12 errors of individual concepts,  but generated no "destructive error” in relational concepts. In conclusion, GT shows a sufficiently reliable translation, and performs admirably and consistently in the three Catholic prayers.

Downloads

Published

2022-10-14
Abstract views: 104 | PDF downloads: 84