Academic profile
Studying the capabilities, limitations and guarantees of AI systems that generate knowledge for people.
OEG / 2026-27
This research line studies how AI systems transform knowledge sources into answers used by people. It is organised around capabilities, limitations and guarantees, with a particular focus on safety and reliability as models, data and tools evolve.
Multilingual knowledge verbalisation and temporal reasoning.
Source relevance, omission bias and risks in generated knowledge.
Safety and reliability throughout the knowledge-generation lifecycle.
A Spanish curriculum-aligned question-answer dataset combining official examination questions and LLM-generated material. The contribution covers dataset design, cognitive criteria and educational evaluation.
An institutional study of question generation and evaluation from academic documents. The contribution covers the methodology, implementation and quantitative and qualitative analysis of the pilot.
A Spanish adaptation of WebNLG for generating text from RDF triples, evaluating efficient language models with in-context learning and partial fine-tuning.
AI and retrieval-augmented generation for analysing long documents through semantic segmentation, knowledge graphs and question answering.
Analysis of European healthcare procurement using natural language processing, knowledge graphs and AI, including the EU Contract Hub platform.
An open Spanish educational question-answer dataset aligned with official curricula and designed for evaluating question generation systems.
Source code and anonymised pilot data for reproducing the QuerIA educational question generation and evaluation experiments.