The Agricultural University of Athens (AUA; Greek: Γεωπονικό Πανεπιστήμιο Αθηνών) is the third oldest university in Greece. Since 1920, it has made contributions to Greek agricultural and economic development, by conducting basic and applied research in Agricultural Science and Technology.The university is situated in the neighborhood of Votanikos, on a 25-hectare green campus that straddles both sides of the historic Iera Odos (the Sacred Way of antiquity), close to the Acropolis, at the heart of the ancient Olive Grove.
Fisher-led area-based fisheries measures require ecological and socio-economic evidence that is scientifically robust, locally legitimate and usable for policy design. The AMORGORAMA initiative on Amorgos Island, central Aegean Sea, addressed this through a fisher-led quadruple-helix partnership among fishers, scientists, NGOs and public authorities, formalized in a 2021 Memorandum of Understanding. This study combined year-round ecological monitoring of small-scale fisheries with a parallel full-census socio-economic survey of the island's 21 active vessels to inform fisher-proposed areas later established as Fisheries Restricted Areas through Presidential Decree P.D. 73/2025. Standardized fishing with gillnets, trammel nets and longlines recorded 3181 individuals from 123 species. Biodiversity peaked in spring, while density was highest in summer, reflecting reproduction and recruitment dynamics. Gear type structured the catch assemblage, highlighting technical measures as adaptive management levers. Length-frequency analyses relative to Minimum Conservation Reference Sizes and L50 showed regulatory misalignments, while conservation-priority taxa occurred at low abundance, indicating exposure to SSF gears. The socio-economic survey quantified fleet dependence and short-run costs of the measures, about EUR 4355 per vessel per year, and specified a transparent compensation option to support an equitable transition. Preliminary ecological and socio-economic results were submitted to the Directorate for Fisheries in December 2023 and considered by the Fisheries Council before formal FRA establishment. The full dataset provides a pre-designation ecological and socio-economic baseline for adaptive monitoring, ecosystem-based fisheries management and future OECM-relevant assessment.
Financial institutions require robust document access control mechanisms that balance security with transparency and explainability. Traditional classification systems often operate as black boxes, failing to provide justifications for access-control decisions. This work presents a novel explainable AI multi-agent recommender system for financial document sensitivity classification that addresses critical ethical concerns in AI-powered decision-making. We fine-tuned three state-of-the-art models—FinBERT, BERT-base-uncased, and GPT-4.1-mini—on a custom-labeled Financial PhraseBank dataset with four sensitivity levels: Public, Internal, Confidential, and Restricted. These fine-tuned models serve as specialized AI agents within a multi-agent architecture orchestrated by GPT-5.1, a large reasoning model operating in zero-shot mode. The orchestrator synthesizes agent predictions and generates natural language recommendations that justify classification decisions. Our agentic AI multi-agent recommender system achieves 83.71
Addressing the instability issues of anthocyanins, the present work studied the encapsulation of anthocyanin rich eggplant peel extract in powders formed by oven drying using gum arabic alone or in mixtures with high and low methoxyl pectins. Moreover, the powders were incorporated in gelatine gels, in order to investigate their exploitation in food formulations as natural colour and antioxidant sources. Despite the heat sensitive nature of anthocyanins, the oven process resulted in great yield (75.31-87 %) and encapsulation efficiency (79.88-94.72 %) values. The powders shared similar moisture content (0.35-0.51 %) and solubility (63.15-63.69 %) with their bulk, tapped and particle densities and porosity varying from 0.67 to 0.75 g/cm(3), 0.70-0.82 g/cm(3), 1.18-2.51 g/cm(3) and 40.7-65.6 %, respectively. They had very good flowability (Carr Index: similar to 5), low cohesiveness (Hausner ratio similar to 1.06), good wettability (<2 s) and their dispersibility ranged from 40 to 53.4 %. Redness was close to +30 as expected due to the presence of anthocyanins. The phenolics content of the powders (0.089-0.124 mg GAE/0.5 mL) was lower than that of the initial extract (0.243 mg GAE/0.5 mL) whereas their antioxidant activity was 40 %. Regarding the wall material, the incorporation of pectins led to powders with increased porosity and particle density, better wettability and worst dispersibility. FT-IR suggested possible polysaccharide-polysaccharide and polysaccharide-extract interactions. No anthocyanins were detected in the gels. However, their AA ranged from 15.46 to 53.79 %. Overall, the present work can be a first step in the modulation of anthocyanin loaded powders for future use as encapsulating, antioxidant and colour agents in food matrices.
The globalization of food supply chains and the proliferation of heterogeneous, multimodal data streams have increased the complexity of food safety management. Limited semantic reasoning capacity, poor data interoperability, and lagging or untimely hazard detection are increasingly constraining traditional governance frameworks, which rely on static databases and fragmented information flows. Knowledge graphs (KGs) provide structured, semantically linked representations of multi-source data, enabling cross-domain integration and causal inference. On the other hand, large language models (LLMs), excel at processing and generating natural language, allowing for automated extraction of safety-critical entities and relationships from unstructured sources. This review presents the latest advances in KG-LLM integration for food safety, highlighting their complementary roles in multimodal knowledge representation, dynamic knowledge fusion, trustworthy inference, and automated construction. Meanwhile, representative applications in expert knowledge mining, intelligent decision-support systems, and consumer-facing services are reviewed, demonstrating a shift from reactive, data-driven responses to proactive risk assessment. Predictably, cross-modal alignment, regulatory compliance, and ethical governance will provide theoretical support for building intelligent, efficient, and interpretable food safety supervision frameworks.
Salinity poses a major threat to agriculture and food security globally. The salinization of soil and water is further deteriorating the pressure that climate change puts on the agrifood sector. Regions that are prone to salinity are reporting significant yield reductions and are coping with suboptimal agricultural production. One such region is the Middle East and North Africa (MENA). MENA constitutes one of the most climate sensitive regions of the world and agriculture is severely hindered by salinity. Despite the extensive research on salinity management in MENA, literature lacks region-wide assessments that could be used for the development of implementable and governance-informed management frameworks. The aim of the present study is to assess the impact of salinity in the countries of MENA, present measures for the mitigation and adaptation to salinity, and facilitate the development of a holistic framework for the management of soil and water salinization. Mitigation and adaptation measures for salinity in MENA include soil, water, and fertilization management, crop and agricultural diversification, breeding and genetic tools, and novel technologies and nature-based solutions. Despite the availability of measures and strategies that could significantly benefit the region in managing salinity, effective and efficient governance is necessary for the successful implementation of any holistic salinity-related policy.
