A Preliminary Study on a Prompt Feedback Loop Structure Based on Human–AI Interaction in Literary Translation

View at publisher Summarise with VIRAT
Authors: Seonhwa LEE
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This study aims to explore the applicability of a stepwise procedural model, the “prompt feedback loop,” which focuses on the interaction between human translators and generative artificial intelligence in literary translation. Rather than treating the output of generative AI as a fixed product, this study conceptualizes translation as an interactive process in which the translator’s judgments and instructions are repeatedly incorporated, and seeks to structure and analyze this process. For this purpose, Kawabata Yasunari’s short short story “Umbrella” (雨傘) was chosen. A four-stage feedback loop was applied to its Japanese–Korean translation. In the first loop, we focused on refining the linguistic foundation to ensure grammatical and lexical accuracy. The second loop addressed the cultural context and the atmospheric qualities of the scenes. The third loop aimed to achieve literary expression by replicating the stylistic rhythm and the author’s distinctive sensibility. In the final loop, the translator intervened directly to fine-tune stylistic choices and enhance the overall completeness of the translated text. The analysis shows that improvements in accuracy and readability were most prominent in the early stages. In contrast, higher-level elements such as emotional conveyance and stylistic representation were progressively reinforced in later stages. In particular, when stylistic guidelines were explicitly specified, literary features such as sentence rhythm, restrained expression, and the handling of narrative pauses became more clearly articulated. Although this study is limited to a preliminary analysis of a single literary work, it suggests the potential of modeling the translation process in generative AI–assisted literary translation and of redefining the translator's role as an interaction designer.

Fields of Study & Subjects

Translation Studies and Practices Natural Language Processing Techniques RNA and protein synthesis mechanisms

Publication details

Journal / source
The Japanese Language Association Of Korea
Volume
87
Pages
67–95
Publication date
26 March 2026
Language
EN