Research on the accuracy and style preservation of a multilingual artificial intelligence translation system based on deep learning in English literary translation

The paper assesses the effectiveness of the multilingual artificial intelligence translation systems which are cutting-edge technologies that translate speech or text between several languages using artificial intelligence while preserving accuracy, meaning, and cultural nuances in translating literary texts and reaffirms the advantages and the inadvertent pitfalls. Sophisticated direct translation is identifiable in these systems, but they struggle with translating formal, comparative, and idiomatic connotations; so essential to literary translation. NMT has been used in many ways and has a great performance, especially in obtaining direct translation without idiomatic, neutral, and far more abstract concepts such as sentiment, difficult grammar, and the actual sense of the word in a metonymic context. Transparency, a primary indicator of translation feasibility, is still unsatisfactory in particular with regards to abstract and poetic texts. Thus, the stylistic equivalence, which is critical for literary translation, proves an 85%-like rate for composing the proper tone, voice intonation, and emotion timbre. Even though contextual embeddings have been developed, careful tracking of such features as shifts in the emotional tone or pacing can need a human’s input to be accurate. The only issue is to consider cultural properties, including idioms and language references when translating. Although there have been important advances, translating literary text is still an open task with challenging problems in abstract, cultural and stylistic cases.