A Comparative Study on Technical Translation Output and Literary Translation Output of Google Translate

Abstract

Machine Translation, with the advance of technology and the growing need for translation, has become very popular all over the world. Google Translate (GT) is an online neural machine translation service, which supports over 100 languages and has more than 200 million users daily. However, the translation quality of some texts that have been translated by GT still remains controversial. Therefore, this paper aims to identify the area Google Translate performs better, in translating literary or technical text from English to Sinhalese. Three literary texts: the poem Daffodils by William Wordsworth, an extract from Earnest Hemingway’s, Indian Camp, and an extract from Garcia Lorca’s drama House of Bernarda Alba, and three technical texts: a tender notice, an abstract and a paragraph from an informative article were selected as source texts to be translated. After these texts were translated by GT the outputs were compared with their original texts. The presentable quality of the translations was evaluated based on the faithfulness to the content and style. Further, the quality of the target language was also measured with regard to syntactic, morphological and semantic aspects. The result shows that from the translations of the given texts, the technical documents were observed to be more faithful to the original texts and of presentable quality, while the literary translations demonstrated several inaccurate outputs, thus require a considerable amount of post editing. Especially the translation of the poem has many errors since the sentence structure in poetry differs from other texts and it is written in figurative language including numerous connotations. Compared to the translation of the poem, the other two literary translations do not show a drastic difference semantically. Findings reveal that GT is more applicable for technical translations rather than literary translations which require post editing