Published in the Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 5: Industry Track), EACL 2026, Rabat, Morocco.
The paper introduces PharmaQA.IT, an Italian extractive question-answering dataset of 861 expert-checked question-answer pairs built from official medicine leaflets (RCP) of the Italian Medicines Agency, using a semi-automatic pipeline with a multimodal LLM over 1,077 leaflets. Tests on open and proprietary LLMs show that open models come close to closed ones when combined with chunking and retrieval.
Authors: Kamyar Zeinalipour, Andrea Zugarini, Asya Zanollo, Leonardo Rigutini.
Read the paper | Code | Models and data
All the lab's language releases are listed in Language models and datasets.