Workers increasingly reassess their positions and many ultimately decide to leave their job, leading to the so-called Great Resignation, which revives a longstanding scholarly interest in the phenomena of attrition. This study aims at exploring literature on attrition and workers’ intentions to leave, by conducting a large‑scale text mining analysis. Drawing on a set of over four thousand articles published until 2025, the application of topic modelling with BERTopic algorithm led to the identification of around 60 topics. The findings demonstrate a wide discussion on such phenomena, highly fragmented yet interconnected landscape: themes span leadership styles, organizational and HRM practices, workplace risks and violence, psychological contract breach, public service motivation, career development, corporate social responsibility, burnout and emotional exhaustion, work-family dynamics, and the profound impact of Covid‑19. The convergence of HR management, nursing, organizational studies, and psychology reveals the complexity of workforce instability. The drivers of attrition operate simultaneously across multiple layers. This underscores the necessity of adopting a holistic lens, translating the interdependencies into actionable insights within the employee employer relationships. By mapping the thematic landscape of attrition, this study offers a first attempt of actionable insights to support companies in designing more informed and effective organizational practices.

Large Language Models (LLMs) are increasingly integrated into innovation and collaborative knowledge management processes. While they demonstrate strong capabilities in text generation and reasoning, their role in transforming tacit expertise into structured, explicit knowledge remains underexplored. This paper proposes a Tacit-to-Explicit Knowledge Detection and Transformation Module designed to formalize experiential expert statements into measurable and reusable knowledge artifacts. The architecture combines rule-based linguistic classification with a structured transformation pipeline that converts subjective inputs into operational procedures. The system was evaluated using four datasets reflecting controlled, natural language, and cross-domain expert contexts. Experimental results show robust performance in domain-aligned conditions, achieving up to 84.2% classification accuracy, while revealing limitations in detecting implicitly expressed tacit knowledge. A representative industrial maintenance use case demonstrates the practical applicability of the approach. The findings confirm the feasibility of LLM-assisted knowledge formalization for collaborative innovation and highlight the need for hybrid semantic and Human–AI reasoning mechanisms to improve robustness and governance.

Hydrogen is widely recognized as a cornerstone of the energy transition, and European policymakers have identified it as a strategic instrument to advance continental decarbonization and long-term energy security objectives. Despite strong policy momentum, however, the sector is struggling to scale. Hydrogen markets are affected by a classic chicken-and-egg problem, in which mutual dependencies among production, infrastructure, and demand hinder large-scale investment deployment. Against this background, the aim of this study is to generate new insights into the structural interdependencies shaping the emerging cross-border hydrogen initiatives. To address this objective, the study draws on ecosystem literature, adopting the ecosystem-as-structure perspective. Using an exploratory qualitative multiple-case design, the analysis proceeds in two steps: first, six hydrogen ecosystems are reconstructed based on their structural features; second, a cross-case comparison is conducted to identify recurring patterns. The findings show that ecosystems are structured around infrastructure-centered configurations characterized by strong sequential alignment of activities and a central role of midstream functions. Consistent with the literature on emergent ecosystems, value propositions appear broad, roles are not yet fully stabilized, and interaction mechanisms remain closely tied to the material sequencing of activities. The cross-border dimension further shapes ecosystems structure, with infrastructure operators assuming a central role in ensuring continuity across jurisdictions. By extending ecosystem research beyond predominantly digital and platform based contexts, this study contributes to a more nuanced understanding of how ecosystem structures are assembled in infrastructure-intensive and transnational energy systems.

Romania's agricultural sector faces a structural gap: despite significant productive potential, no functional extension system exists at national level. The European Union (EU) Agricultural Knowledge and Innovation System (AKIS) framework under Common Agricultural
Policy (CAP) 2023–2027 creates both the obligation and the opportunity to address this absence. Existing European models offer only partial guidance for contexts where no legacy system exists to build upon.
This paper argues that establishing agricultural extension in Romania is not primarily an agronomic challenge, but a digital transformation one. Scalable knowledge transfer from research to farm level increasingly depends on digital infrastructure — satellite-based crop monitoring, IoT sensor networks, and AI-driven decision support - that falls within the core competence of technical universities rather than agronomic institutions. Yet technical universities remain absent or peripheral in existing AKIS governance structures across Europe.
Drawing on a comparative analysis of European AKIS models and on direct research collaboration with the University of Georgia Cooperative Extension, this paper proposes a framework that positions technical universities as structural actors within a university-based extension consortium. The framework operates through three interconnected components: a nationally coordinating technical laboratory responsible for platform architecture and data governance, regional technical nodes embedded across Romania's development regions for in-situ data collection, and a unified EPlatform designed to integrate heterogeneous data streams, from satellite imagery and sensor networks to agronomic records, market data, and farmer-generated inputs, into actionable, AI-driven recommendations.

The proposed model adapts the core logic of the WUR digital extension architecture to a fragmented institutional context, offering a realistic and EU-aligned path toward operationalizing AKIS where none currently exists — and demonstrating that agricultural extension, reimagined through digital transformation, can function as the operational delivery layer of a national knowledge and innovation system.