EFRA

Project team:

• Tony Lindgren, Associate Professor, DSV, SU
• John Pavlopoulos, Affiliated Researcher, DSV, SU
• Korbinian Randl, PhD student, DSV, SU

Partners:

• Agroknow, Greece
• National Research Council (Cnr), Italy
• Wageningen Food Safety Research Institute, Netherlands
• Maize, Italy
• Agrivi, Croatia
• Rainno, Greece
• SGS Digicomply, Romania
• Moy park, UK

Project period: 2023-01-01 to 2025-12-31

Funding: European Union

Description

EFRA aspires to develop the first analytics-enabled, secure-by-design, green data space for AI-enabled food risk prevention. The project will explore how extreme data mining, aggregation and analytics may address major scientific, economic and societal challenges associated with the safety and quality of the food that European consumers eat. The mission of the project is to support EU’s global leadership in the digital-led industry transition from reaction to food risk prevention.

The project will explore three use-cases:

1. Risk predictions for poultry pathogens: EFRA will monitor and predict the presence of the most usual pathogens in poultry farms, using lab test and environmental data coming directly from farm sensors and publicly available data sources.
2. Food-safety Optimal Pesticide Use: By leveraging both aggregated data from at-the-farm sensors and publicly available data sources, EFRA will deploy and train AI models, for optimal pesticide use recommendations.
3. Informing Regulatory Decisions with Food Risk Intelligence: To identify potential emerging food safety risks, EFRA will use extensive set of regulatory data and mining algorithms, while different individuals, companies and organisations will upload relevant data using the EFRA tools.

To this end it will utilize NLP, machine learning and explainable methods for knowledge discovery from hetrogenious data sources (including multilingual data). The intention is to build a novel data management and analytics framework, based on three pillars: (1) Data hub – searchable data integration of heterogenous data, (2) Analytic tools – used together with the data hub to distill and combine useful insights regarding food safety from available data, and (3) Data and Analytics marketplace – a single resource for stakeholders in food safety to interact with both data and predictions methods for food safety. The final product will be a set of methods and tools for integrating massive and heterogeneous food safety related data sources, and a set of predictive models for learning from these data sources, with emphasis on interpretability and explainability of the models’ rationale for the predictions.

The main work packages of the project are:

1. Extreme Data Discovery & Mining
   
2. Data Analytics & AI Prediction Models
   
3. EFRA Green Data & Analysis Infrastructure
   
4. Use-cases, Focus Groups & Performance Experiments
   

People

Tony Lindgren, Associate Professor

explainability, predictive maintanance

John Pavlopoulos

machine learning for natural language processing, applications to healthcare and education

Korbinian Randl

text classification, explainability in NLP, food risk prediction