Can Pikachu evolve in Turkish?


Creative Commons License

Kuşgöz M. T.

STRIDON: Journal of Studies in Translation and Interpreting, cilt.6, sa.1, ss.27-53, 2026 (Scopus)

Özet

This study replicates and extends the study of Moreno García and Mangiron (2024) and investi­gates the efficacy of GPT-5 in transcreating Pokémon names into Turkish within the Prompt-based Conversational Machine Translation (PCMT) framework. By applying a novel typology across five case studies, the research evaluates the model’s performance in creativity, semantic integrity, cul­tural resonance and phonological plausibility. The findings reveal that while GPT-5 demonstrates high semantic agility and morphological inventiveness, it exhibits vulnerabilities regarding Turkish morphophonological constraints and sociolinguistic safety. These limitations necessitate a manda­tory layer of “constrained transcreation”, refuting the notion of fully autonomous AI generation. Consequently, the study suggests that the integration of LLMs does not signal the obsolescence of the human translator, but rather the evolution of their role from a linguistic mediator to a strategic “prompt architect” and a culturally and sociolinguistically sensitive assessor of the AI output. Ulti­mately, this research validates a symbiotic workflow where computational power is tempered by the indispensable cultural and structural oversight of a human co-agent.