CULTURAL AND EMOTIONAL EQUIVALENCE IN LITERARY TRANSLATION: AN EXPLAINABLE AI FRAMEWORK FOR CULTURAL-MARKER ANALYSIS

This study develops and evaluates an interpretable system for quantifying the preservation of emotional impact in literary translation. Its relevance follows from the gap between rapidly advancing multilingual sentiment analysis and the limited availability of methods that jointly model multidimensional affect, dominant emotion, and culturally marked context. The proposed hybrid approach combines eight Plutchik emotions, the valence-arousal-dominance space, and cultural-marker analysis. Emotional Distance (ED), Preservation of Dominant Emotion (PDE), and Cultural Adaptation (K) are integrated into an Emotional Quality Assessment score (EQA). A Python prototype was evaluated in more than twenty experiments on passages from 1984 and Harry Potter and the Philosopher's Stone, using two Russian translations of each novel. Mean EQA scores were 0.71 and 0.64 for the translations by V. Golyshev and L. Bershidsky, and 0.78 and 0.72 for the Rosman and M. Spivak versions. EQA correlated with three experts at r = 0.82, while dominant-emotion accuracy reached 79%. The method supports preliminary comparison but does not replace expert judgement, particularly for irony, implicit affect, and stylistic deviation.