Transformasi penerjemahan mesin (mt) dan tinjauan dampak ai terhadap etika dan masa depan penerjemah

Abstract

This article outlines the history and evolution of Machine Translation (MT) as a response to the need for efficient cross-language communication in the era of globalization. Starting with Warren Weaver's mathematical idea in 1949 and the formal Georgetown-IBM experiment in 1954, MT initially adhered to the Rule-Based Approach, which relied on syntax and semantics manually engineered by linguists. Limitations in language variation then prompted a transition to Statistical Machine Translation (SMT) in the 1990s, which used probability models based on parallel data corpora, yielding more flexible translations but demanding massive amounts of data. The modern era of machine translation is dominated by Neural Machine Translation (NMT), which emerged in the mid-2010s thanks to advancements in deep learning and AI's transformer architecture. NMT uses artificial neural networks to process the meaning of a sentence as a whole, producing translations that are far more natural and contextual than their predecessors, as seen in popular systems like Google Translate and DeepL. NMT also allows for specific stylistic adjustments and domain-specific translation. Although NMT has achieved remarkable progress in terms of accuracy and efficiency, challenges remain, including high computational power requirements for training, limitations in handling idioms or highly specific cultural contexts, and issues of bias inherited from the training data. Nevertheless, the development of AI and the potential for multimodal data integration promise an increasingly accurate and contextual future for MT, one that is moving ever closer to the quality of human translation.