Applications and Projects
The core competence about innovation through applied Artificial Intelligence results in the involvement of the members of the SAG group into a variety of projects, experimental studies or evaluation campaigns. These are opportunities to assess AI technology against real-world challenging problems within industrial scenarios with specific technological and business key performance indexes (KPIs). Results correspond to cutting-edge systems that combine deep learning, natural language processing, and semantic technologies to solve real-world problems — from healthcare and cultural heritage to education, sustainability, and national security. Each project stems from close collaboration with industry, government, and academic partners, allowing the group to translate foundational AI research into scalable, ethical, and impactful solutions.
Brand Reputation
Partners: ENEL, RAI
AI-driven systems for large-scale opinion mining and sentiment analysis are very important for industry being able to materialize crucial information about a company’s market. Among the different domains in which SAG carried out research (e.g. tourism), the media and energy industry represent two noticeable comprehensive examples. For RAI (the Italian National Radio Television company), a platform analyzing Twitter data to support marketing and public engagement has been designed, optimized and integrated into a cloud based service oriented architecture for real time Twitter analysis. In the energy domain, ENEL (the Italian National Industry for Electrical Energy), a zero-shot learning brand reputation framework has been designed and optimized for modeling, extracting and monitoring public perception in the energy sector. The resulting AI services enable real-time accurate processing of open Web sources and Brand Reputation evaluation and tracking.
Epidemic Intelligence
Partner: Istituto Superiore di Sanità – ISS
We design AI systems for the early detection and monitoring of emerging health threats by leveraging semantic analysis, topic discovery, and large language models. Our methods allow the automatic classification of health-related news and social signals, supporting real-time situational awareness. Special focus is placed on zero-shot classification and the interpretability of language model outputs for public health analysts.
Big Data for Tourism
Partner: Unioncamere, ISNART
Our work supports tourism planning and destination management through predictive analytics and semantic processing of large-scale data sources. We build AI systems to analyze user reviews, identify emerging trends, and assess destination reputation. By integrating statistical indicators with social perception and sustainability metrics, our research enables data-driven tourism policies and long-term strategic vision.
Advanced OCR for High Quality IE
Partner: ACI
We processed over 80 million scanned documents from vehicle registration archives (1920–1990), applying deep learning-based OCR and information extraction techniques. Our work enabled high-accuracy data validation for license plate records, combining computer vision and semantic models for structured knowledge generation at scale.
Predictive Analytics
Partners: ISNART, VEAS
We designed AI systems for predictive modeling in complex industrial and tourism contexts. In collaboration with VEAS in Norway, we applied deep neural architectures for anomaly detection in wastewater treatment plants. With ISNART, we contributed to the Stendhal project, forecasting tourism dynamics and supporting sustainable policy planning.
HRI – Human-Robot Interaction
Partner: Campus Bio-Medico di Roma
We develop multimodal and adaptive interfaces for robot-assisted rehabilitation, integrating natural language understanding, computer vision, and affective computing. Our systems allow physical and cognitive interaction between patients and robotic platforms, supporting personalized therapy and real-time emotional feedback. Visual grounding and sensor fusion are key components of our HRI architectures.
Open Source Intelligence
Partner: Ministero dell’Interno
Our platform “Re4CT – Revealer for Crime Tracking” applies natural language processing and machine learning to support intelligence workflows. It enables automated analysis of large text corpora, semantic indexing, and threat pattern detection. This system has been integrated in institutional settings for enhanced public security and OSINT capabilities.
AI for Education
Partners: Meta Group, Scuola IAD
We are building LLM-powered agents capable of autonomously collecting educational materials, performing advanced information retrieval, summarization, and concept aggregation. Our vision is to enable AI tutors that support students and instructors through intelligent access to didactic content, aligned with ethical and transparent learning frameworks.
KnowledgeKwnoledge Integration for Nuclear Decommissioning
Partner: IAEA – International Atomic Energy Agency
We apply natural language processing and semantic search technologies to support knowledge extraction and document management in the nuclear energy sector. Our systems enable the classification, indexing, and retrieval of regulatory and technical documents through domain-specific language models. This supports transparency, traceability, and decision-making in sensitive and safety-critical processes.
AI for Sustainability Reporting
We design AI systems that support sustainable development goals through explainable and ethically grounded automation. Our models assist organizations in extracting environmental, social, and governance indicators from documents, ensuring compliance, transparency, and strategic decision-making in ESG reporting.
AI for Genomics
Partner: Policlinico Tor Vergata – PTV
Our group develops deep learning models for transcriptomic data, with the goal of improving cancer diagnostics and biomarker discovery. We integrate explainable AI to ensure the interpretability of predictions and reduce model complexity without loss of performance. This work bridges machine learning, precision medicine, and bioinformatics for next-generation medical AI.
Deep Learning and Cultural Heritage
Partner: Faculty of Humanities, University of Rome Tor Vergata
We apply natural language understanding and multimodal AI to the cultural sector. Our work focuses on building natural language interfaces to query structured knowledge bases in the domain of performing arts, such as historical theater archives, enabling intuitive and intelligent access to cultural heritage.