From User Preferences to Optimization Constraints Using Large Language Models
Manuela Sanguinetti, Alessandra Perniciano, Luca Zedda, Andrea Loddo, Cecilia Di Ruberto, Maurizio Atzori
TL;DR
The paper addresses translating natural language user preferences into energy optimization constraints for home appliances within Italian Renewable Energy Communities. It evaluates Italian-language LLMs under zero-shot, one-shot, and few-shot prompting using XML-tagged prompts and a two-step annotation pipeline to produce a formal constraint representation FR(c) with variables such as $s_t$ and $h_t$. The main contributions include establishing a baseline for this task, publicly releasing the pilot dataset and code, and providing insights on best practices and limitations, notably the need for domain-specific fine-tuning. This work informs NLP-driven approaches to energy management and user-centric optimization in REC contexts, highlighting practical challenges and directions for future research.
Abstract
This work explores using Large Language Models (LLMs) to translate user preferences into energy optimization constraints for home appliances. We describe a task where natural language user utterances are converted into formal constraints for smart appliances, within the broader context of a renewable energy community (REC) and in the Italian scenario. We evaluate the effectiveness of various LLMs currently available for Italian in translating these preferences resorting to classical zero-shot, one-shot, and few-shot learning settings, using a pilot dataset of Italian user requests paired with corresponding formal constraint representation. Our contributions include establishing a baseline performance for this task, publicly releasing the dataset and code for further research, and providing insights on observed best practices and limitations of LLMs in this particular domain