EKWE JONAL

This study investigates a critical challenge in machine translation: while Neural Machine Translation (NMT) has significantly enhanced fluency and contextual accuracy, it continues to struggle with rendering cultural and idiomatic expressions, particularly in low-resource languages. Using qualitative content analysis, this research compares NMT with Rule-Based Machine Translation (RBMT) and Statistical Machine Translation (SMT) through two case studies: Chinua Achebe’s Things Fall Apart (translated from English to Igbo) and Luke 15:11–12 (translated from Greek and English toIgbo). RBMT typically produces rigid, literal translations, while SMT introduces probabilistic refinements, yet both approaches fail to fully capture cultural depth. Although NMT offers greater fluidity and contextual awareness, it remains inadequate in conveying culturally embedded meanings, particularly in biblical narratives with profound moral and social implications. The research findings underscore NMT’s limitations in handling complex idiomatic and cultural elements, despite its improved readability. By empirically assessing NMT’s performance in literary and scriptural contexts, this research contributes to computational linguistics and translation studies. It provides valuable insights for computational linguists, translation scholars, AI developers, and policymakers, emphasizing the need for refining machine translation models to enhance their ability to translate culturally rich texts accurately and contextually in low-resource languages.

Open Full Text (PDF)