From Literal to Cultural: Advancing Machine Translation with Sociolinguistic and Sentiment-Based Approaches

This study examines the capabilities of machine translation (MT models) in preserving cultural subtleties, specifically when translating content rich in idiomatic expressions, emotional tones, and cultural context. As MT models become increasingly integrated into international communication, concerns have been raised about their ability to accurately capture these nuances. Using movie subtitles as a reference point, the research explores the distinctions between manual and machine translations, analyzing how well MT systems maintain the integrity of the original content. Bias detection methods, cross language consistency testing, and cultural sensitivity scenarios are employed to evaluate and compare AI generated translations against human outputs. The study also uses corpus-based and sentiment analysis, along with sociolinguistic evaluations, to identify the limitations of existing MT models. Survey and user feedback provide additional insights, reinforcing the need for more culturally aware translation algorithms. By leveraging advanced NLP frameworks like Transformer and BERT models, this research suggests adaptations that prioritize cultural nuance over literal accuracy. Ultimately, the goal is to propose solutions that enhance both the accuracy and cultural sensitivity of machine translations, facilitating more effective cross language communication while reducing reliance on time consuming manual translation.