For five days, Delft University of Technology (TU Delft) became the meeting point for the international scientific community dedicated to information retrieval, bringing together around 500 experts to discuss search and recommendation models, the integration of language models, and the challenges associated with evaluating and improving the efficiency of these systems. INESC TEC took part in the 48th edition of the European Conference on Information Retrieval (ECIR 2026) with a team of researchers presenting several scientific papers and organising a workshop.
INESC TEC is already recognised for the scientific contributions across multiple areas of Artificial Intelligence (AI), and at ECIR 2026 it stood out with the presentation of eight scientific papers, the organisation of an international workshop, and the participation of Ricardo Campos, INESC TEC researcher, in the conference’s scientific coordination as Full Paper Chair.
“Participation in ECIR 2026 reinforces INESC TEC’s position in a highly competitive and rapidly evolving scientific field. It is a globally recognised conference, and the largest in Europe in information retrieval, where some of the most significant advances in this domain are discussed. INESC TEC’s presence, with multiple accepted papers and contribution to the scientific organisation, reflects the consistency and impact of the work our team has been developing in this research line,” emphasised the researcher, who is also a lecturer at the University of Beira Interior (UBI).
The scientific papers presented showed how the institution and the partners (the University of Porto and UBI) have been exploring the intersection between information retrieval and Natural Language Processing, with applications ranging from the extraction and structuring of information from public administration documents to the detection of disinformation and multimodal search.
One of the highlights the CitiLink project, a platform that transforms minutes from municipal meetings into structured and searchable information, contributing to greater transparency and civic participation. Developed in collaboration with several Portuguese municipalities, this project also led to CitiLink-Minutes, a new annotated dataset containing 120 municipal minutes, more than one million tokens and approximately 45,000 annotations, as well as the MiNER model for automatic metadata extraction, which first identifies the most important sections and then extracts specific data such as names, dates or locations. These contributions provide a solid foundation for future research in information retrieval and natural language processing.
Another important contribution was ClaimPT, an annotated dataset in Portuguese built from news produced by the LUSA agency, designed to support the development of automated fact-checking systems. By identifying verifiable claims in journalistic texts, this resource contributes to advancing tools to reduce misinformation, addressing the lack of high-quality data available in Portuguese for this area.
In the field of multimodal search, the researchers presented the ImageSeek system, combining text and images to improve search results and enabling users to find images based on natural language descriptions. Alongside this, the pt-image-ir-dataset was introduced to support systems such as ImageSeek, also addressing the shortage of resources in this area for the Portuguese language.
“These works demonstrate how research in information retrieval and natural language processing, particularly in Portuguese, can translate into tools with direct societal impact, namely in organising and accessing information, tackling misinformation, and improving search systems. They contribute not only to advancing the state of the art but also to establishing solid foundations for the development of solutions with real-world impact,” the researcher commented.
In addition to the scientific presentations, INESC TEC was also involved in organising the 9th edition of the Text2Story 2026 workshop, dedicated to the automatic extraction and understanding of narratives. The event brought together international researchers and included scientific sessions, invited talks and awards for the best papers, helping to energise this scientific community – which over nine editions has already involved more than 500 researchers.
“Text2Story has established itself as a key forum for discussing approaches to the automatic extraction and understanding of narratives within the ECIR context, bringing together an increasingly consolidated international community. The continuity of this initiative over the past nine years reflects the growing relevance of the topic and how it has evolved alongside the field, incorporating new challenges and approaches to the extraction, representation and analysis of narratives from text,” explained Ricardo Campos.
INESC TEC’s participation in ECIR 2026 reflects the institution’s growing international recognition in the fields of information retrieval and AI, strengthening the presence within leading scientific networks. The next edition of ECIR will take place in Southampton between 21 and 25 March 2027.
News, current topics, curiosities and so much more about INESC TEC and its community!
