In this episode of the ELOQUENCE podcast series, the conversation focused on the rapidly evolving field of speech technology, multilingual AI, and the role of Europe in developing language technologies that are accurate, inclusive, ethical, and aligned with European values.
The guest was Alessio Brutti, Scientific Coordinator of the ELOQUENCE project and Head of the Speech Technology Lab at Fondazione Bruno Kessler. With many years of experience in speech recognition, digital audio processing, multimodal signal processing, and AI-driven communication systems, Alessio shared insights into both the scientific foundations and the broader strategic goals of the project.
Alessio began by reflecting on his journey into speech and audio technologies. His background in telecommunication engineering and signal processing led him to work on European research projects, later followed by a PhD and a long-term research career at FBK. Over the years, the field has changed dramatically. What was once mostly signal processing with some machine learning has now been transformed by deep learning, large language models, and the rapid growth of AI-based tools.
The episode also highlighted the work of the Speech Technology Lab at FBK, one of the institute’s longest-standing research groups. Its core expertise lies in automatic speech recognition, but the team also works on spoken language understanding, speech enhancement, speaker identification, diarisation, and related areas. Today, much of their research focuses on low-resource languages and domains, as well as solutions that can be adapted to specific contexts where general-purpose tools are not enough.
A key topic of the conversation was Work Package 4 in the ELOQUENCE project, which focuses on speech technologies that go beyond pure accuracy. Alessio emphasised two especially important dimensions: multilinguality and ethics. Many European languages can be considered low-resource compared to globally dominant languages such as English or Chinese, which means that high-quality language technologies are not equally available to all European citizens and organisations. ELOQUENCE aims to address this by supporting the development of tools that work across different European languages and contexts.
The ethical dimension is equally important. The project seeks to ensure that models are privacy-compliant, bias-aware, and aligned with European values. To achieve this, technical development must go hand in hand with expertise from partners working on ethics, law, and human rights.
The discussion also explored the importance of the human touch in AI-driven communication. While today’s tools are powerful, interactions with AI systems can still be frustrating when machines fail to understand users properly. Alessio pointed to personalisation, context awareness, and continual learning as key directions for making AI systems more useful, comfortable, and trustworthy, especially in sensitive domains such as healthcare call centres.
Looking ahead, Alessio expressed support for more open science in AI and speech technology. Rather than relying only on closed tools developed by major technology companies, he sees value in European research labs and smaller companies joining forces to develop open models and better understand the systems they use.
The episode offered a clear view of how ELOQUENCE contributes to Europe’s broader language technology ecosystem: by advancing multilingual, trustworthy, efficient, and human-centred AI for real-world communication.
Listen to the full episode here.
