The Analysis of the Impact of Machine Translation on Expressing the Cultural Value of Literary Works from the Cross Cultural Perspective

Abstract

Cultural fidelity, semantic retention, and linguistic fluency in machine translation (MT) systems has become a problem as MT systems tend to fail at preserving cultural fidelity, cultural oversimplification, and idiomatic and metaphorical language. Taking a hybrid, human-machine translation models approach to address these challenges across numerous literatures is the subject of this study. The hybrid model achieved significant quantitative and qualitative advancements as found in the study. The hybrid model quantitatively exceeds other systems in 0.92 BLEU, 0.90 METEOR, 0.94 linguistic fluency, and 0.88 cultural appropriateness, 27.4%, 26.7%, and 45.9% improvements respectively. The qualitative results suggested that the hybrid model maintained 'Very High' scores in idiomatic expressions, metaphorical language and cultural essence, outperforming standalone MT systems. MT often applied such misrepresentation to the misrepresented key cultural values of xenia (Confucian ideals in Chinese culture), ‘asabiyyah (tribal solidarity in Arab culture), Ubuntu’s ethos (in African culture) and existentialism (in Western culture) greatly oversimplifying ideas like xiao (filial piety), ‘asabiyyah (tribal unity), and Sartre’s existentialism. These differences were preserved in hybrid models, which also ensured cultural and philosophical fidelity in translations by reflecting, for example, Ubuntu’s communal harmony and Ibn Khaldun’s principles of social justice. The superior hybrid model is characterized by this integration of semantic precision, cultural depth and stylistic naturalness. Through this study, we show how hybrid models can help bridge the gap between computational efficiency and cultural sensitivity in translation studies.