Tradução automática e inteligência artificial: uma análise historiográfica e perspectivas futuras

Abstract

Over the past decades, the evolution of digital technologies has significantly reshaped the field of translation, especially through the emergence of machine translation systems and, more recently, generative artificial intelligences (AIs). This study investigates the intervention of AI in translation by adopting a mixed methodological approach qualitative, quantitative, exploratory, and historiographical to examine the technical, historical, and social aspects of this transformation. The research is structured in three stages: a literature review of the historical development of machine translation (from Rule-Based to Neural and Generative systems), a comparative analysis of translation engines (such as Google Translate, DeepL, Amazon Translate, and GPT-based models), and a prospective study on future trends and challenges. Results show that although Large Language Models (LLMs) like GPT-4 outperform traditional Neural Machine Translation (NMT) systems in contextual fluency and adaptability, while still exhibit translationese traits and face limitations in under-resourced languages. The study concludes that the future of translation lies in hybrid models that combine the generative power of LLMs with domain-specific fine-tuning, alongside the continued relevance of human translators in ensuring quality, ethics, and cultural nuance.