Natural Language Processing in Healthcare
1st AIxIA Summer School on Artificial Intelligence for Healthcare
07-11 July 2025, Trento, Italy
1st AIxIA Summer School on Artificial Intelligence for Healthcare
07-11 July 2025, Trento, Italy
Natural Language Processing and Digital Health: Challenges and Opportunities
M. Dragoni
Integrating Natural Language Processing (NLP) and Digital Health transforms how healthcare data is analyzed, interpreted, and utilized. From clinical decision support and patient engagement to predictive analytics and automated documentation, NLP-driven technologies offer immense potential to enhance efficiency and improve patient outcomes. However, several challenges, such as data privacy concerns, bias in AI models, domain-specific language complexities, and regulatory constraints, must be addressed to ensure the responsible deployment of these technologies. This lecture will explore the opportunities and challenges of NLP and Digital Health, highlighting real-world applications, emerging trends, and ongoing research efforts. We will discuss how advancements in deep learning, large language models, and federated learning are shaping the future of NLP in healthcare while considering ethical and practical limitations. Attendees will gain insights into the current landscape and future directions, equipping them with a deeper understanding of how NLP can be leveraged to drive innovation in digital health.
Large Language Models and Healthcare
L. Sanna
Recent advancements in artificial intelligence, particularly Large Language Models (LLMs), are transforming the healthcare landscape. These models have demonstrated remarkable potential in clinical decision support, medical research, patient communication, and administrative efficiency. This lecture explores the capabilities of LLMs in healthcare, addressing their applications in diagnosing diseases, generating medical documentation, and enhancing telemedicine services. We will also discuss the ethical considerations, biases, and regulatory challenges associated with AI in medicine. This session will provide insights into how LLMs can augment medical professionals while ensuring patient safety and data security by examining real-world case studies and emerging trends.
Building a pre-validated conversational agent for Digital Health
S. Magnolini
Conversational agents are transforming Digital Health by providing scalable, intelligent, and user-friendly interactions for patient engagement, diagnostics, and mental health support. However, ensuring their reliability and compliance with healthcare standards requires rigorous validation before deployment. This lecture explores a framework for building pre-validated conversational agents tailored for Digital Health applications. We will cover key aspects such as dataset curation, model training with domain-specific constraints, ethical considerations, regulatory compliance (e.g., HIPAA, GDPR), and real-world validation methodologies. Developers can streamline regulatory approval and enhance user trust by integrating pre-validation techniques, such as synthetic data generation, expert-in-the-loop evaluation, and safety monitoring. Attendees will gain insights into best practices for designing AI-driven health assistants that are effective and safe for clinical and patient use.
NLP and Psychology: a look to the future
L. Sanna
Natural Language Processing (NLP) and psychology are increasingly intersecting, shaping the future of human-computer interaction, mental health assessment, and cognitive research. This lecture explores how cutting-edge NLP models revolutionize psychological studies, from analyzing emotions in text to detecting mental health disorders through linguistic patterns. We will discuss the ethical implications, challenges of bias, and the potential for AI-driven therapy and personalized interventions. Looking ahead, we envision a future where NLP not only aids psychologists but also enhances our understanding of human cognition, communication, and well-being.
Practical session - Hands-on session: Building an FAQ agent LLM-based
P. Bellan
Participants will be divided into mixed inter-disciplinary groups and asked to answer a practical research question, building an LLM-based pre-validated conversational agent.
MAURO DRAGONI, PhD
Intelligent Digital Agents Research Group, Fondazione Bruno Kessler, Trento
Mauro Dragoni is a research scientist at Fondazione Bruno Kessler where he is the Head of the Intelligent Digital Agents research unit (IDA). His main research topics concern knowledge management, information retrieval, and machine learning by focusing on developing real-world prototypes as the outcome of his research activities. He has been involved in a number of international research projects at the national and international levels. He co-authored over 160 scientific publications in international journals, conferences, and workshops.
LEONARDO SANNA, PhD
Intelligent Digital Agents Research Group, Fondazione Bruno Kessler, Trento
Leonardo Sanna is a researcher specializing in advancing conversational AI technologies for healthcare applications. With a background in linguistics and communication, his work focuses on the trustworthy integration of Large Language Models (LLMs) into multi-agent systems. His research investigates the pragmatic capabilities of LLMs, with a particular focus on their ability to interpret implicit meaning. His work also involves the development of novel evaluation metrics and methodologies to assess the performance and reliability of text generation.
SIMONE MAGNOLINI, PhD
Intelligent Digital Agents Research Group, Fondazione Bruno Kessler, Trento
Simone Magnolini has been a researcher at Fondazione Bruno Kessler (FBK) since 2017, specializing in natural language processing (NLP) and conversational agents, particularly in healthcare applications. With a background in computer science and engineering, his work explores multi-agent systems, misunderstanding detection and classification, as well as synthetic data generation. His research delves into the inner workings of large language models (LLMs), focusing on their ability to improve performance without task-specific training. More recently, his work has also addressed bias detection in the outputs of LLMs.
PATRIZIO BELLAN, PhD
Intelligent Digital Agents Research Group, Fondazione Bruno Kessler, Trento
Patrizio Bellan is a researcher focusing on Natural Language Processing and Generative AI. With a background in Computer Science and Cognitive Science, his research interests include multi-agent systems, retrieval-augmented generation, and knowledge extraction methods in low-resource settings. He explores techniques for integrating personality into large language models to improve their performance on complex and underspecified reasoning tasks.