The Duality of Machine Translation: A Critical Analysis of Literal and Free Translation Approaches in AI Systems

Abstract

The fast development of machine translation (MT) based on Artificial Intelligence (AI) as a system of neural machine translation (NMT) has transformed the realm of cross-linguistic communication. Such systems usually work within the range of literal (form-based) translation strategies and free (meaning-based) ones. The paper will critically discuss the inherent drawbacks of the literal and free translation methods as applied by AI. Based on a literature review, it contends that even literal AI translation is prone to generating grammatically aberrant, culturally insensitive, and semantically erroneous output, whereas free AI translation is prone to bring out uninformed distortion, loss of critical nuance, and bias. The paper concludes that the main problem is not the decision between the two extremes, but the fact that the AI is not yet capable of providing the complex contextual, cultural and pragmatic judgment in order to perform faithful and functional translations. The consequences to the users, developers and translation profession are addressed.

Keywords: machine translation, AI translation, literal translation, free translation, neural machine translation, translation errors, cultural bias