We present an AI-powered tool developed within the Horizon Europe-funded BI4E (Boosting INGENIUM for Excellence) project, which aims to foster new, interdisciplinary research collaborations across a university network. A key challenge addressed by BI4E is the difficulty of forming interdisciplinary and cross-university research teams, for instance for applying to competitive funding. The CONNECT platform was designed to fill this gap by providing a semantic search engine that allows researchers from different universities to discover potential collaborators on the basis of their scientific publications. At the core of CONNECT there is a transformer-based language model, used to encode both funding call descriptions and researcher publication metadata into high-dimensional vector representations. The metadata are collected from the Scopus database using ORCID or Scopus identifiers. When queried with the topic of competitive fundings, CONNECT returns a list of researchers whose publications semantically aligns with the input query. In addition, CONNECT integrates an explainability mechanism derived from the attention matrix mechanism, allowing users to understand which parts of the input text most influenced the match. The aim of CONNECT is to support the strategic goals of the BI4E project by offering an institutional view of research domains, facilitating industry-university matchmaking, and democratizing access to collaboration opportunities. Since its launch in December 2023, CONNECT has already enabled the formation of dozens of new research teams within the INGENIUM European University Alliance, validating its potential as a catalyst for research excellence and institutional transformation.
Enabling research collaboration with AI: the BI4E experience with large language models
Gianluca Amato;Marcello Costantini;Francesca Ferri;Maria Chiara Meo;Francesca Scozzari
;
2025-01-01
Abstract
We present an AI-powered tool developed within the Horizon Europe-funded BI4E (Boosting INGENIUM for Excellence) project, which aims to foster new, interdisciplinary research collaborations across a university network. A key challenge addressed by BI4E is the difficulty of forming interdisciplinary and cross-university research teams, for instance for applying to competitive funding. The CONNECT platform was designed to fill this gap by providing a semantic search engine that allows researchers from different universities to discover potential collaborators on the basis of their scientific publications. At the core of CONNECT there is a transformer-based language model, used to encode both funding call descriptions and researcher publication metadata into high-dimensional vector representations. The metadata are collected from the Scopus database using ORCID or Scopus identifiers. When queried with the topic of competitive fundings, CONNECT returns a list of researchers whose publications semantically aligns with the input query. In addition, CONNECT integrates an explainability mechanism derived from the attention matrix mechanism, allowing users to understand which parts of the input text most influenced the match. The aim of CONNECT is to support the strategic goals of the BI4E project by offering an institutional view of research domains, facilitating industry-university matchmaking, and democratizing access to collaboration opportunities. Since its launch in December 2023, CONNECT has already enabled the formation of dozens of new research teams within the INGENIUM European University Alliance, validating its potential as a catalyst for research excellence and institutional transformation.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
