Antonio Castaldo
PhD Researcher in Artificial Intelligence
About Me
I am a PhD researcher in Artificial Intelligence and Machine Translation at University of Pisa, where I explore the intersection of language, creativity, and computational models. My research focuses on the use of large language models (LLMs) for creative text translation, with an emphasis on literary and non-standard language domains.
I am especially interested in responsible and adaptive training strategies for machine translation, human-centered evaluation of post-editing workflows, and the development of tools that support translators' creative agency. I advocate for MT systems that are not just efficient, but also aligned with the nuances of human expression—particularly in queer and culturally rich texts.
Currently, I am working on a thesis that investigates how LLMs can be fine-tuned and evaluated for creativity in literary translation, integrating post-editing data and multi-faceted creativity metrics. My aim is to challenge conventional MT paradigms and contribute to a more inclusive, linguistically sensitive future for AI-driven translation technologies.
Research Interests
Machine Translation & Language Models
- Large Language Models (LLMs) for Machine Translation
- Literary Translation and Creative Text Generation
- Domain Adaptation and Dataset Curation for Low-Resource Literary Texts
Human-Centered Evaluation
- Post-editing, Editing Effort, and Human-Centered Evaluation
- Creativity Metrics (humor, wordplay, style, surprise)
Creative Translation & Evaluation
- Literary Translation and Non-Standard Language
- Translator-Centered Post-Editing Workflows
- Creativity-Aware Evaluation for Style, Humor, and Wordplay
Recent News
- Paper and poster presented at MT Summit 2025, Geneva.
- Presented project at PhD Summer School in AI & Democracy, Pisa.
- Two papers accepted at MT Summit 2025.
- Released UniOr PET, an online post-editing tool.