Archivio Istituzionale della Ricerca - Università degli Studi di Pavia
Not a member yet
135341 research outputs found
Sort by
Evolutionary paths of Global Value Chains (GVCs) for SMEs: A conceptual framework
No full textExternal factors such as geopolitical tensions, protectionist pressures, and technological evolution are posing significant challenges to multinational enterprises (MNEs) and to smaller firms in a wide variety of sectors and nations. The aim of this conceptual chapter is to devote special attention to
identifying the new dynamics and evolutionary paths of global value chains (GVCs) in the context of small- and medium-sized enterprises (SMEs). Moreover, it aims to understand, more specifically, the complex context of international business affecting and challenging SMEs involved in GVCs
and propose some future directions about how they are responding to the uncertainty generated by the increasing dynamicity of GVCs. The findings are systematized within a conceptual model. This research contributes to the research streams studying the relational dynamics within the GVCs. Regarding managerial implications, this study will enable SMEs to make better-informed decisions regarding their future directions and business models, strengthening their capacity to respond to abrupt shocks within the GVC of lead firms
Evil plants and perilous waters: Science and technology in the Po Valley (18th and 19th centuries)
No full textIn the 18th and 19th centuries, scholars and technicians became more and more interested in issues related to public health, such as food security, good hygiene in towns and the countryside, and the fight against ‘paludism’. Through the analysis of correspondence, reports, projects, and technical texts, this chapter examines some case studies from the Po Valley related to public health implementation plans in northern Italy, an area particularly productive in terms of agriculture and animal husbandry, but also rich of waters and swampy areas. The goal of the chapter is to discuss the more or less ‘sustainable’ attempts to reconcile the agricultural sector, public health, and the environment in a key period for the evolution of science and technology in Italy
GLI ULTERIORI ARRETRAMENTI DELLA TUTELA PENALE IN MATERIA DI TERRORISMO E DI INCOLUMITÀ PUBBLICA
No full textIndividual-based foundation of SIR-type epidemic models: mean-field limit and large-time behaviour
No full textWe introduce a kinetic framework for modelling the time evolution of the statistical distributions of the population densities in the three compartments of susceptible, infectious and recovered individuals, under epidemic spreading driven by susceptible-infectious interactions. The model is based on a system of Boltzmann-type equations describing binary interactions between susceptible and infectious individuals, supplemented with linear redistribution operators that account for recovery and reinfection dynamics. The mean values of the kinetic system recover a SIR-type model with reinfection, where the macroscopic parameters are explicitly derived from the underlying microscopic interaction rules. In the grazing collision regime, the Boltzmann system can be approximated by a system of coupled Fokker–Planck equations. This limit allows for a more tractable analysis of the dynamics, including the large-time behaviour of the population densities. In this context, we rigorously prove the convergence to equilibrium of the resulting mean-field system in a suitable Sobolev space by means of the so-called energy distance. The analysis reveals the dissipative structure of the dynamics and the role of the interaction terms in driving the system towards a stable equilibrium configuration. These results provide a multi-scale perspective connecting kinetic theory with classical epidemic models
Sviluppo e validazione di strumenti di intelligenza artificiale generativa applicati alle malattie genetiche rare.
No full textL’intelligenza artificiale (IA) generativa è emersa negli ultimi anni come una tecnologia rivoluzionaria con ampie applicazioni nel campo scientifico, tra cui la ricerca biomedica e l’assistenza sanitaria. L’IA generativa comprende una famiglia di modelli progettati per creare nuovi esempi di dati a partire da dati reali. All’interno dei modelli di IA generativa, i modelli linguistici di grandi dimensioni (LLM) hanno riscosso successo grazie alla loro capacità di generare testi simili a quelli umani e di eseguire compiti di ragionamento complessi sfruttando reti neurali profonde addestrate su enormi set di dati. La loro rapida adozione in tutti i settori biomedici ha consentito progressi nella ricerca in diversi ambiti. Nel contesto della bioinformatica e delle malattie rare, gli LLM hanno introdotto nuove opportunità per analizzare dati biologici complessi e accelerare nuove scoperte. Esempi di applicazioni includono l’interpretazione di sequenze genetiche, la previsione della struttura delle proteine e la formulazione di farmaci attraverso la generazione di nuove molecole. Inoltre, gli LLM assistono i flussi di lavoro bioinformatici, recuperano e sintetizzano la letteratura scientifica in genomica e proteomica, consentendo una annotazione più efficiente e precisa delle varianti genomiche e delle informazioni fenotipiche. Nonostante questi progressi, l’adozione dei modelli LLM in contesti biomedici deve affrontare diverse sfide. La riproducibilità rimane un problema a causa della stocasticità intrinseca dei modelli generativi. L’interpretabilità degli LLM è limitata dalla loro natura di black box, che complica la convalida e presenta problemi di responsabilità. Anche la privacy dei dati è un problema, poiché l’addestramento di tali modelli richiede spesso grandi set di dati contenenti informazioni sensibili sui pazienti. Questo lavoro esplora due applicazioni degli LLM nel campo della genetica: VarChat, uno strumento per creare sintesi delle varianti genetiche, e PhenoXtract, uno strumento per l’estrazione di fenotipi standardizzati da testi clinici. Queste applicazioni dimostrano come gli LLM possano supportare la genomica e la bioinformatica clinica, evidenziando anche la necessità di strategie ibride per superare i limiti intrinseci dei modelli. VarChat è una piattaforma progettata per estrarre e sintetizzare risultati rilevanti dalla letteratura scientifica su una variante genomica. Con un framework di Retrieval-Augmented Generation (RAG), VarChat recupera le pubblicazioni relative alla variante, seleziona i chunks più rilevanti e li fornisce come input all’LLM. Il modello genera quindi riassunti supportati da riferimenti bibliografici. Questo approccio garantisce l’accuratezza delle informazioni riducendo al contempo il tempo necessario per l’interpretazione della letteratura sulle varianti, offrendo a clinici e ricercatori uno strumento efficiente e affidabile. PhenoXtract è uno strumento per l’estrazione di informazioni fenotipiche da dati clinici non strutturati, che introduce una metodologia ibrida che integra gli LLM con gli embedding di Knowledge Graphs (KG) e la Human Phenotype Ontology (HPO), con l’obiettivo di mappare i termini estratti alle voci HPO standardizzate. VarChat dimostra che un LLM basato su un approccio ottimizzato di recupero della letteratura è in grado di sintetizzare in modo efficiente le informazioni curate sulle varianti e migliorare i flussi di lavoro di interpretazione delle varianti. PhenoXtract dimostra che le strategie ibride che combinano LLM e ontologie producono risultati accurati e standardizzati. Applicati al contesto delle malattie rare, i due tools illustrano esempi complementari per l’applicazione dell’IA generativa. Questo lavoro contribuisce all’applicazione dell’IA generativa alla bioinformatica, sottolineando l’importanza di combinare l’IA con conoscenze specifiche del settore per risultati significativi.Generative Artificial Intelligence (AI) has emerged in recent years as a revolutionary technology with broad applications in science, included in biomedical research and healthcare. Generative AI comprises a family of models designed to create new instances of data from real data. Within Generative AI models, Large Language Models (LLMs) have gained success due to their ability to generate human-like text and perform complex reasoning tasks by leveraging deep neural networks trained on massive datasets. Their rapid adoption across biomedical domains has enabled research progress in several areas. In the context of bioinformatics and rare diseases, LLMs and related generative models have introduced new opportunities to analyze complex biological data and facilitate new discoveries. Examples of applications include the interpretation of genetic sequences, protein structure prediction, drug discovery through novel molecule generation. Furthermore, LLMs assist bioinformatics workflows by retrieving and summarizing scientific literature in genomics and proteomics, enabling more efficient and precise reporting of genomic variants and phenotypic information. Despite this progress, the adoption of LLMs in biomedical contexts faces several challenges. Reproducibility remains an issue due to the inherent stochasticity of generative models. Interpretability is limited by their black-box nature, complicating validation and presenting accountability issues. Data privacy is also an issue, as training such models often requires large data sets containing sensitive patient information. These challenges underline the importance of ethical control, domain-specific validation, and hybrid approaches that combine generative models with biomedical knowledge. This work explores two applications of LLMs within genetics: Var Chat, a tool for automating comprehensive genetic variant summaries, and PhenoXtract, a tool for the extraction of standardized phenotypic descriptions from clinical texts. Together, these applications demonstrate how LLMs can support genomics and clinical bioinformatics, also showing the need for hybrid strategies to overcome inherent model limitations. VarChat is a platform designed specifically to search, extract, and synthesize relevant results from the scientific literature on a genomic variant. By implementing a Retrieval-Augmented Generation (RAG) framework, VarChat retrieves variant-related publications, selects the most relevant text chunks, and provides them as input to the LLM. The model then generates reference-supported summaries that integrate the literature evidence into a comprehensive report. This approach ensures information accuracy while reducing the time required for variant literature curation and interpretation, offering clinicians and researchers an efficient, trustworthy tool for analyzing genetic variation. PhenoXtract is a tool for extracting phenotypic information from unstructured clinical data. PhenoXtract introduces a hybrid methodology that integrates LLMs with knowledge graph (KG) embeddings and the Human Phenotype Ontology (HPO), with the aim of mapping extracted terms to standardized HPO entries. VarChat demonstrates that a LLM powered by an optimized literature retrieval strategy approach can efficiently synthesize variant cu rated information and improve variant interpretation workflows. PhenoXtract shows that hybrid strategies combining LLMs with ontologies produce accurate, standardized outputs. Applied to rare disease context, the two tools illustrate complementary strategies for applying generative AI in variant interpretation and clinical text analysis. This work contributes to the application of generative AI to bioinformatics, highlighting the importance of combining generative AI with domain-specific knowledge resources to achieve reliable and clinically meaningful results
Arctic diel and circadian acoustic pattern of Orcas, Fin, and Humpback whales revealed by deep learning from two months of continuous recordings
No full textIn the northern Norwegian fjords, orcas, fin and humpback whales gather each winter to feed on herring, overlapping with intense human activities such as fishing and whale watching. To assess how anthropophony and geophony influence their acoustic behavior, we conducted two months of continuous passive acoustic monitoring in Kvænangen fjord during winter 2022–2023. Whale vocalizations were automatically detected using a deep learning framework based on YOLOv5, enabling quantification of species-specific acoustic presence and activity. Ambient noise was estimated from power spectral density. Low- and high-noise conditions were identified for geophony and anthropophony using a three-step filtering procedure. Model performances were evaluated under various noise conditions to ensure robust and consistent detection accuracy. Analyzes were then performed to characterize diel, circadian and daily patterns of acoustic activity. All three species were detected nearly continuously, with orcas activity peaking in November. Acoustic patterns were strongly influenced by noise: orcas and fin whales were less vocally active with increasing anthropophony (ρ< −0.31, p < 0.05), while humpback whales showed a time-dependent response, increasing vocal activity on short timescales (p < 0.01) but decreasing over longer periods (ρ = −0.33, p = 0.008). Geophony was associated with reduced acoustic presence for all three species on a daily basis (ρ< −0.34, p < 0.01), suggesting changes in spatial distribution or vocal behavior. Positive correlations between orcas and humpback whales vocal behavior indicated potential concurrent feeding. These findings revealed species- and timescale-specific acoustic responses to noise and illustrate how deep-learning can enhance ecoacoustic monitoring
Greek New Comedy Beyond Menander
No full textWhile Menander has long stood as the emblematic figure of Greek New Comedy, recent scholarship has increasingly challenged this Menandrocentric view by re-examining the broader landscape of comic production in the late-Classic and Hellenistic period. This volume brings to the fore the fragmentary remains of Menander’s contemporaries—such as Philemon, Diphilus, Apollodorus and others—offering new insights into their character typologies, linguistic style, and political engagement. Through philological, literary, and reception-oriented approaches, the contributions in this book reevaluate long-standing assumptions about the genre. The volume not only recontextualizes New Comedy within its historical and literary frameworks but also sheds light on its impact on Roman comedy. By expanding the canon beyond Menander, this study provides classicists, philologists, and literary historians with a more nuanced and comprehensive understanding of the genre's complexity and cultural significance
The antimicrobial effect and pain control of ozonized gel versus gaseous ozone in the management of primary molars caries: randomized clinical trial with in vivo microbiological evaluation
No full textBackground: Pediatric dentistry has been radically revolutionized in recent years with the atraumatic approach (atraumatic restorative treatment (ART)), which is able to preserve the vitality of the tooth, with a higher attention to the child’s well-being and emotional comfort. In this scenario, the use of ozone with its decontaminating power and bacterial load reduction, can play a decisive role in the application of the principles of selective caries removal (SCR). The aim of this study was to evaluate and compare the efficacy of ozonized gel (GeliO3) and gaseous ozone (healOzone X4) by analysing colony forming unit (CFU) microbial load of Streptococcus mutans (S. mutans) after caries removal, together with patient sensitivity and compliance. Methods: 16 patients aged 4–12 years were enrolled and randomly assigned to gel or gas treatment. Baseline dental and periodontal indices, such as Gingival Index (GI), Plaque Index (PI), International Caries Detection and Assessment System (ICDAS), and Basic Erosive Wear Examination, were assessed. After manual caries removal with an excavator, microbiological samples were collected with a sterile paper cone at T0 (before treatment), T1 (after ozone application for 30 seconds), and T2 (after an additional 30 seconds application). Schiff Air Index (SAI) and Face, Legs, Activity, Cry and Consolability (FLACC) were recorded at T0 and at T2 to evaluate patients’ compliance and dental sensitivity of the clinical procedure. Data underwent statistical analysis (significance: p 0.05). SAI and FLACC showed no significant inter-and intragroup differences at any time point (p > 0.05). Conclusions: GeliO3 and healOzone X4 demonstrated comparable antimicrobial effect against S. mutans. The procedures had no emotional impact on paediatric patients. Clinical Trial Registration: clinicaltrials.gov (NCT06641323)