Emotion Profiling in LLM-Based Literary Translation: Systematic Shifts Across MT and Post-Editing
arXiv:2606.10113v1 Announce Type: new Abstract: This paper investigates whether LLM translations exhibit identifiable emotional profiles and how post-editing reshapes them toward human-like norms. We compare LLM translations of Margaret Atwood's Oryx and Crake with their post-edited versions and a human translation, using a large-scale corpus of contemporary Italian science-fiction as a baseline. We examine emotion through lexicon-based and multilingual modeling, conducting a fine-grained analysis of emotional variation across systems. We find that MT systems introduce model-specific and stati
The proliferation of LLMs in translation and creative applications necessitates a deeper understanding of their nuanced outputs, particularly in fields like literary translation where emotional fidelity is paramount.
This research provides critical insights into the inherent biases and systematic emotional shifts introduced by LLMs, informing the development of more sophisticated and 'human-aligned' AI translation tools.
The understanding of LLM emotional profiling allows for targeted post-editing and fine-tuning strategies to mitigate unintended emotional distortions in machine-translated texts, especially in sensitive domains.
- · Literary translators
- · LLM developers
- · Post-editing tools
- · Multilingual content creators
- · Generic MT models
- · Purely automated translation workflows
Increased focus on emotional intelligence and stylistic nuance in LLM fine-tuning for creative and sensitive text generation.
Development of specialized LLMs or modules specifically designed to preserve or manipulate emotional profiles in translation and content creation.
Ethical guidelines and regulatory frameworks emerging for controlling emotional manipulation or distortion by AI in translated works and public communication.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.CL