The Centre for Research in Information and Communication Technologies (CITIC) at the University of A Coruña (UDC) has strengthened its international profile following its participation in the 47th European Conference on Information Retrieval (ECIR 2025), one of the most prestigious and demanding scientific events in the field of information retrieval. Held from April 6 to 10 in the Italian city of Lucca, this edition had an acceptance rate of just 23%, making CITIC’s presence with five accepted papers a remarkable achievement.

Six members of the Information Retrieval Lab (IRlab) research group presented five papers addressing key challenges at the intersection of artificial intelligence, search systems, and digital well-being. From the use of generative models to enhance information retrieval through contextual synonyms, to the development of more efficient and explainable systems for hate speech detection, CITIC’s proposals stand out for their innovation and focus on socially impactful issues.

In the paper titled «Leveraging Retrieval-Augmented Generation for Keyphrase Synonym Suggestion», researchers Jorge Gabín and Javier Parapar propose an approach based on Retrieval-Augmented Generation (RAG) to enhance keyword synonym suggestions in search systems. This method expands queries with alternative concepts using language models, reducing unsuccessful searches and improving search engine accuracy, thereby optimizing the user experience.

Paloma Piot and Javier Parapar presented «Towards Efficient and Explainable Hate Speech Detection via Model Distillation», a compact language model that, through distillation techniques and Chain-of-Thought reasoning, improves both classification and explainability in the detection of hate speech on social media. This proposal aligns with Green AI principles by offering a more efficient and sustainable solution.

David Otero, Javier Parapar, and Álvaro Barreiro authored «Towards Reliable Testing for Multiple Information Retrieval System Comparisons», a study proposing a new framework for reliable testing in the comparison of multiple information retrieval systems, ensuring more accurate and fair evaluations in the field.

The contribution «MindWell: A Demonstration of an Early Risk Detection System for Mental Health», presented by Eliseo Bao, Anxo Pérez, and Javier Parapar, showcased a demo of an open-source conversational agent based on a large language model (LLM). Aimed at professional use, the system enables early detection of depressive symptoms by contextually analyzing social media posts, providing explanations aligned with clinical frameworks such as the Beck Depression Inventory-II (BDI-II).

Lastly, in  «eRisk: Presentación de una Tarea de CLEF sobre Detección Temprana de Riesgos en Internet», researchers Javier Parapar and Anxo Pérez presented the new “eRisk” task within the international CLEF framework, which aims to advance methodologies and metrics for early risk detection on the internet, particularly in contexts related to health and digital safety.

CITIC’s participation in ECIR 2025 not only reinforces its leadership in the field of information retrieval, but also reflects its research community’s commitment to scientific excellence, responsible artificial intelligence, and impactful knowledge transfer to society.