2026
Journal article
Open Access
Enhancing token boundary detection in disfluent speech
Srivastava Manu, Ferro Marcello, Pirrelli Vito, Coro Gianpaolo
This paper presents an open-source Automatic Speech Recognition (ASR) pipeline optimised for disfluent Italian read speech, designed to enhance both transcription accuracy and token boundary precision in low-resource settings. The study aims to address the difficulty that conventional ASR systems face in capturing the temporal irregularities of disfluent reading, which are crucial for psycholinguistic and clinical analyses of fluency. Building upon the WhisperX framework, the proposed system replaces the neural Voice Activity Detection module with an energy-based segmentation algorithm designed to preserve prosodic cues such as pauses and hesitations. A dual-alignment strategy integrates two complementary phoneme-level ASR models to correct onset–offset asymmetries, while a bias-compensation post-processing step mitigates systematic timing errors. Evaluation on the READLET (child read speech) and CLIPS (adult read speech) corpora shows consistent improvements over baseline systems, confirming enhanced robustness in boundary detection and transcription under disfluent conditions. The results demonstrate that the proposed architecture provides a general, language-independent framework for accurate alignment and disfluency-aware ASR. The approach can support downstream analyses of reading fluency and speech planning, contributing to both computational linguistics and clinical speech research.Source: INTELLIGENT SYSTEMS WITH APPLICATIONS, vol. 29
DOI: 10.1016/j.iswa.2025.200614
Project(s): READLET
Metrics:
Srivastava Manu, Ferro Marcello, Pirrelli Vito, Coro Gianpaolo
This paper presents an open-source Automatic Speech Recognition (ASR) pipeline optimised for disfluent Italian read speech, designed to enhance both transcription accuracy and token boundary precision in low-resource settings. The study aims to address the difficulty that conventional ASR systems face in capturing the temporal irregularities of disfluent reading, which are crucial for psycholinguistic and clinical analyses of fluency. Building upon the WhisperX framework, the proposed system replaces the neural Voice Activity Detection module with an energy-based segmentation algorithm designed to preserve prosodic cues such as pauses and hesitations. A dual-alignment strategy integrates two complementary phoneme-level ASR models to correct onset–offset asymmetries, while a bias-compensation post-processing step mitigates systematic timing errors. Evaluation on the READLET (child read speech) and CLIPS (adult read speech) corpora shows consistent improvements over baseline systems, confirming enhanced robustness in boundary detection and transcription under disfluent conditions. The results demonstrate that the proposed architecture provides a general, language-independent framework for accurate alignment and disfluency-aware ASR. The approach can support downstream analyses of reading fluency and speech planning, contributing to both computational linguistics and clinical speech research.Source: INTELLIGENT SYSTEMS WITH APPLICATIONS, vol. 29
DOI: 10.1016/j.iswa.2025.200614
Project(s): READLET
Metrics:
See at:
CNR IRIS 
| www.sciencedirect.com | CNR IRIS 
2026
Journal article
Open Access
Detecting fishing areas from navigation radar detector data
Coro Gianpaolo, Bove Pasquale, Ellenbroek Anton
Detecting global fishing activity is essential for sustainable ocean governance, yet systems based on vessel-transmitted information, such as Automatic Identification System (AIS) and Vessel Monitoring Systems, are limited by access issues, coverage gaps, and the inability to detect non-cooperative vessels. To overcome these issues, this paper presents Point-to-Fishing (P2F), an AI-driven workflow to detect fishing areas and estimate fishing hours from Navigation Radar Detector (NRD) data of satellite or terrestrial systems, complemented with currents and bathymetry data from Copernicus and GEBCO. P2F integrates analytical components based on statistical analysis, machine learning, and deep learning to conduct vessel behaviour analysis, spatial feature extraction (vessel abundance, recurrence, current-driven interpolation, and bathymetric suitability), and anomaly detection. The workflow operates effectively with or without vessel identifiers, enabling the detection of fishing areas in data-sparse or AIS-denied regions, even using one satellite only. P2F is validated on data covering the North Sea, the Western Norwegian Sea, and the North Atlantic. The validation cases utilise terrestrial and satellite NRD data alternately, with the Global Fishing Watch fishing effort distributions as a validation reference. P2F achieves a consistent ~75% agreement in relevant fishing area classification and intense-fishing area identification, and ~93% accuracy in total fishing effort estimation.Source: INTERNATIONAL JOURNAL OF DIGITAL EARTH, vol. 19 (issue 1)
DOI: 10.1080/17538947.2026.2617004
Project(s): Accordo di Collaborazione N. 403036 del 25/10/2024 tra la Norwegian Defence Research Establishment (FFI) e il CNR-ISTI.
Metrics:
Coro Gianpaolo, Bove Pasquale, Ellenbroek Anton
Detecting global fishing activity is essential for sustainable ocean governance, yet systems based on vessel-transmitted information, such as Automatic Identification System (AIS) and Vessel Monitoring Systems, are limited by access issues, coverage gaps, and the inability to detect non-cooperative vessels. To overcome these issues, this paper presents Point-to-Fishing (P2F), an AI-driven workflow to detect fishing areas and estimate fishing hours from Navigation Radar Detector (NRD) data of satellite or terrestrial systems, complemented with currents and bathymetry data from Copernicus and GEBCO. P2F integrates analytical components based on statistical analysis, machine learning, and deep learning to conduct vessel behaviour analysis, spatial feature extraction (vessel abundance, recurrence, current-driven interpolation, and bathymetric suitability), and anomaly detection. The workflow operates effectively with or without vessel identifiers, enabling the detection of fishing areas in data-sparse or AIS-denied regions, even using one satellite only. P2F is validated on data covering the North Sea, the Western Norwegian Sea, and the North Atlantic. The validation cases utilise terrestrial and satellite NRD data alternately, with the Global Fishing Watch fishing effort distributions as a validation reference. P2F achieves a consistent ~75% agreement in relevant fishing area classification and intense-fishing area identification, and ~93% accuracy in total fishing effort estimation.Source: INTERNATIONAL JOURNAL OF DIGITAL EARTH, vol. 19 (issue 1)
DOI: 10.1080/17538947.2026.2617004
Project(s): Accordo di Collaborazione N. 403036 del 25/10/2024 tra la Norwegian Defence Research Establishment (FFI) e il CNR-ISTI.
Metrics:
2026
Journal article
Open Access
Estimating species commonness and prevalence through unsupervised methods
Bove Pasquale, Bertini Andrea, Coro Gianpaolo
The prevalence of a species in a given area is crucial for estimating the environmental conditions associated with its subsistence within ecological niche models (ENMs). Prevalence is defined as the proportion of presences relative to the total number of sampled sites, reflecting prior expectation on species commonness or rarity. However, reliable estimation often faces challenges due to limited or biased occurrence data, particularly for rare or poorly monitored species. This work presents a data-driven, multi-species methodology to estimate species prevalence for use in ENMs. It leverages species occurrence records from the Global Biodiversity Information Facility and is entirely unsupervised. It utilises two clustering methods, one deep-learning model, and an ensemble model, plus statistical analysis to classify species commonness and transform classifications into prevalence probabilities. A case study is presented for 161 species living in the Massaciuccoli Lake basin (Tuscany, Italy), a wetland of high biodiversity value and ecological sensitivity. The models classified the species’ prevalence based on observations from other Italian wetland sites, and were evaluated against expert-based assessments. All models achieved high accuracy, with the deep-learning model achieving the highest (~ 81–90%). The proposed methodology is scalable and reproducible and can inform ENMs with objective, robust prevalence estimates.Source: SCIENTIFIC REPORTS, pp. 1-30
DOI: 10.1038/s41598-026-38900-1
Project(s): Italian Integrated Environmental Research Infrastructures System
Metrics:
Bove Pasquale, Bertini Andrea, Coro Gianpaolo
The prevalence of a species in a given area is crucial for estimating the environmental conditions associated with its subsistence within ecological niche models (ENMs). Prevalence is defined as the proportion of presences relative to the total number of sampled sites, reflecting prior expectation on species commonness or rarity. However, reliable estimation often faces challenges due to limited or biased occurrence data, particularly for rare or poorly monitored species. This work presents a data-driven, multi-species methodology to estimate species prevalence for use in ENMs. It leverages species occurrence records from the Global Biodiversity Information Facility and is entirely unsupervised. It utilises two clustering methods, one deep-learning model, and an ensemble model, plus statistical analysis to classify species commonness and transform classifications into prevalence probabilities. A case study is presented for 161 species living in the Massaciuccoli Lake basin (Tuscany, Italy), a wetland of high biodiversity value and ecological sensitivity. The models classified the species’ prevalence based on observations from other Italian wetland sites, and were evaluated against expert-based assessments. All models achieved high accuracy, with the deep-learning model achieving the highest (~ 81–90%). The proposed methodology is scalable and reproducible and can inform ENMs with objective, robust prevalence estimates.Source: SCIENTIFIC REPORTS, pp. 1-30
DOI: 10.1038/s41598-026-38900-1
Project(s): Italian Integrated Environmental Research Infrastructures System
Metrics:
2026
Journal article
Open Access
Interactive story maps for historical musical instruments: a 3D and semantic web tool for cultural heritage preservation
Bartalesi Valentina, Lenzi Emanuele, De Martino Claudio, Coro Gianpaolo
Ancient stringed instruments from the 17th and 18th centuries are essential for global cultural heritage but pose preservation and study challenges due to their fragility and rarity. Digital reproductions and interactive story maps offer new ways to analyse these instruments, providing historical, mechanical, and acoustic insights while aiding modern luthiers in replication. This paper presents a novel methodology and software that integrates 3D models of historical instruments into interactive story maps using Semantic Web technologies. The system enables users to explore 3D models with embedded annotations from Sketchfab and supports the creation of formal narratives enriched with geospatial and multimedia content. Each narrative event is linked to a backend knowledge base (KB) structured with established ontologies. The tool is open-source and web-based, and allows interoperability between its story maps and external KBs such as Wikidata and Europeana. It supports multiple levels of descriptive details, from precise 3D model annotations to broad geographic and temporal references. An evaluation case study featured 25 participants exploring a story map about a 1737 Antonio Stradivari violin, connected to a KB of 1716 triples. The results highlight the tool’s usability and effectiveness in representing the violin’s spatiotemporal and cultural context, showcasing the value of combining 3D models and geospatial data to enhance cultural heritage preservation.Source: INTERNATIONAL JOURNAL OF DATA SCIENCE AND ANALYTICS, vol. 22 (issue 1)
DOI: 10.1007/s41060-025-01014-4
Metrics:
Bartalesi Valentina, Lenzi Emanuele, De Martino Claudio, Coro Gianpaolo
Ancient stringed instruments from the 17th and 18th centuries are essential for global cultural heritage but pose preservation and study challenges due to their fragility and rarity. Digital reproductions and interactive story maps offer new ways to analyse these instruments, providing historical, mechanical, and acoustic insights while aiding modern luthiers in replication. This paper presents a novel methodology and software that integrates 3D models of historical instruments into interactive story maps using Semantic Web technologies. The system enables users to explore 3D models with embedded annotations from Sketchfab and supports the creation of formal narratives enriched with geospatial and multimedia content. Each narrative event is linked to a backend knowledge base (KB) structured with established ontologies. The tool is open-source and web-based, and allows interoperability between its story maps and external KBs such as Wikidata and Europeana. It supports multiple levels of descriptive details, from precise 3D model annotations to broad geographic and temporal references. An evaluation case study featured 25 participants exploring a story map about a 1737 Antonio Stradivari violin, connected to a KB of 1716 triples. The results highlight the tool’s usability and effectiveness in representing the violin’s spatiotemporal and cultural context, showcasing the value of combining 3D models and geospatial data to enhance cultural heritage preservation.Source: INTERNATIONAL JOURNAL OF DATA SCIENCE AND ANALYTICS, vol. 22 (issue 1)
DOI: 10.1007/s41060-025-01014-4
Metrics:
See at:
CNR IRIS | link.springer.com | CNR IRIS | CNR IRIS
2025
Journal article
Open Access
Computing ecosystem risk hotspots: a mediterranean case study
Coro G., Pavirani L., Ellenbroek A.
In ecosystem management, risk assessment quantifies the probability and impact of events and informs on intervention priorities. Analytical models for risk assessment quantify the impact of natural and anthropogenic stressors on ecosystems. Traditional approaches evaluate single stressors, whereas complex models assess cumulative impacts of frequently interacting stressors and offer better accuracy at the expense of low cross-area re-applicability and long implementation times. We introduce a versatile, re-useable, and semi-automated workflow designed for big data-driven ecosystem risk assessment, utilising spatiotemporal data from open repositories. It allows for a flexible definition of the stressors on which the risk under analysis depends. By applying cluster analysis, the workflow identifies different patterns of stressor concurrency, while statistical analysis highlights clusters of stressors likely linked to elevated risk. Ultimately, it generates geospatial risk maps and identifies spatial risk hotspots. The workflow methodology is independent of the geographical area of the application. As a case study, we present risk assessments for the Mediterranean Sea, a region with intense anthropogenic pressures and significant climatic vulnerabilities. We used over 1.1 million open data from 2017 to 2021 and projections to 2050 under the RCP8.5 scenario (a high greenhouse gas emission scenario) at a 0.5°spatial resolution. Data included environmental, oceanographic, biodiversity variables, and manifest and hidden fishing effort distributions. Our workflow identified different types of high-risk hotspots, highlighting different concurrencies of habitat loss, overfishing, hidden fishing, and climate change stressors. High-risk hotspots concentrated in the Western Mediterranean, the Tyrrhenian Sea, the Adriatic Sea, the Strait of Sicily, the Aegean Sea, and eastern Turkey. Our results agreed with an alternative Fuzzy C-means-based method (with a 90% to 96% overlap over the years) and a Bayesian regression model (∼80% overlap). Our Mediterranean risk maps can facilitate the development of management and monitoring strategies, supporting the sustainable development and resilience of coastal zones, and can act as prior knowledge for ecosystem models and spatial plans.Source: ECOLOGICAL INFORMATICS, vol. 85
DOI: 10.1016/j.ecoinf.2024.102918
Project(s): EcoScope
Metrics:
Coro G., Pavirani L., Ellenbroek A.
In ecosystem management, risk assessment quantifies the probability and impact of events and informs on intervention priorities. Analytical models for risk assessment quantify the impact of natural and anthropogenic stressors on ecosystems. Traditional approaches evaluate single stressors, whereas complex models assess cumulative impacts of frequently interacting stressors and offer better accuracy at the expense of low cross-area re-applicability and long implementation times. We introduce a versatile, re-useable, and semi-automated workflow designed for big data-driven ecosystem risk assessment, utilising spatiotemporal data from open repositories. It allows for a flexible definition of the stressors on which the risk under analysis depends. By applying cluster analysis, the workflow identifies different patterns of stressor concurrency, while statistical analysis highlights clusters of stressors likely linked to elevated risk. Ultimately, it generates geospatial risk maps and identifies spatial risk hotspots. The workflow methodology is independent of the geographical area of the application. As a case study, we present risk assessments for the Mediterranean Sea, a region with intense anthropogenic pressures and significant climatic vulnerabilities. We used over 1.1 million open data from 2017 to 2021 and projections to 2050 under the RCP8.5 scenario (a high greenhouse gas emission scenario) at a 0.5°spatial resolution. Data included environmental, oceanographic, biodiversity variables, and manifest and hidden fishing effort distributions. Our workflow identified different types of high-risk hotspots, highlighting different concurrencies of habitat loss, overfishing, hidden fishing, and climate change stressors. High-risk hotspots concentrated in the Western Mediterranean, the Tyrrhenian Sea, the Adriatic Sea, the Strait of Sicily, the Aegean Sea, and eastern Turkey. Our results agreed with an alternative Fuzzy C-means-based method (with a 90% to 96% overlap over the years) and a Bayesian regression model (∼80% overlap). Our Mediterranean risk maps can facilitate the development of management and monitoring strategies, supporting the sustainable development and resilience of coastal zones, and can act as prior knowledge for ecosystem models and spatial plans.Source: ECOLOGICAL INFORMATICS, vol. 85
DOI: 10.1016/j.ecoinf.2024.102918
Project(s): EcoScope
Metrics:
See at:
Ecological Informatics | CNR IRIS | www.sciencedirect.com | CNR IRIS
2025
Journal article
Open Access
Gap analysis on the biology of marine fishes across european seas
Kesner-Reyes K., Capuli E. C., Reyes Jr R. B., Jansalin J. G. M., Rius-Barile J., Bactong M., Daskalaki E., Manousi S., Ferrà Carmen, Scarcella G., Coro G., Ordines F., Celie L., Scotti M., Lambert C., Gal G., Palomares M. L., Tsikliras A. C., Dimarchopoulou D.
This review evaluates the current knowledge of essential biological traits (diet, fecundity, maturity, length-weight relationships, spawning, growth, lifespan, and natural mortality) of marine fishes across European and adjacent waters. These traits are crucial for ecosystem modeling and stock assessments. Using data from FishBase, the largest and most comprehensive database on fishes, a gap analysis was performed to identify areas of research focus and the corresponding gaps that require further study. Biological data coverage is strong in the Baltic and North Seas but moderate in the Adriatic, Aegean, Biscay, Celtic, Levantine, and western Mediterranean Seas. Well-documented species include the European conger (Conger conger), thornback ray (Raja clavata), and transparent goby (Aphia minuta) which are reported from all areas. The narrowest knowledge gaps concern length-weight relationships, followed by spawning and growth, while natural mortality and fecundity are the least studied biological characteristics. Regional variations exist, particularly for protected species. Future research should focus on filling gaps by addressing overlooked species (bycatch and discarded species) and traits such as natural mortality and fecundity, with special attention to vulnerable groups like sharks and rays. Expanding biological data coverage will reduce uncertainties in stock assessments and improve ecosystem models, two widely used tools for sustainable fisheries management and marine conservation.Source: REVIEWS IN FISHERIES SCIENCE & AQUACULTURE, pp. 1-22
DOI: 10.1080/23308249.2024.2446806
Project(s): EcoScope, EcoScope
Metrics:
Kesner-Reyes K., Capuli E. C., Reyes Jr R. B., Jansalin J. G. M., Rius-Barile J., Bactong M., Daskalaki E., Manousi S., Ferrà Carmen, Scarcella G., Coro G., Ordines F., Celie L., Scotti M., Lambert C., Gal G., Palomares M. L., Tsikliras A. C., Dimarchopoulou D.
This review evaluates the current knowledge of essential biological traits (diet, fecundity, maturity, length-weight relationships, spawning, growth, lifespan, and natural mortality) of marine fishes across European and adjacent waters. These traits are crucial for ecosystem modeling and stock assessments. Using data from FishBase, the largest and most comprehensive database on fishes, a gap analysis was performed to identify areas of research focus and the corresponding gaps that require further study. Biological data coverage is strong in the Baltic and North Seas but moderate in the Adriatic, Aegean, Biscay, Celtic, Levantine, and western Mediterranean Seas. Well-documented species include the European conger (Conger conger), thornback ray (Raja clavata), and transparent goby (Aphia minuta) which are reported from all areas. The narrowest knowledge gaps concern length-weight relationships, followed by spawning and growth, while natural mortality and fecundity are the least studied biological characteristics. Regional variations exist, particularly for protected species. Future research should focus on filling gaps by addressing overlooked species (bycatch and discarded species) and traits such as natural mortality and fecundity, with special attention to vulnerable groups like sharks and rays. Expanding biological data coverage will reduce uncertainties in stock assessments and improve ecosystem models, two widely used tools for sustainable fisheries management and marine conservation.Source: REVIEWS IN FISHERIES SCIENCE & AQUACULTURE, pp. 1-22
DOI: 10.1080/23308249.2024.2446806
Project(s): EcoScope
, EcoScope Metrics:
See at:
CNR IRIS | www.tandfonline.com | ZENODO | CNR IRIS | CNR IRIS
2025
Journal article
Open Access
Distributed environments for ocean forecasting: the role of cloud computing
Ciliberti S., Coro G.
Cloud computing offers an opportunity to innovate traditional methods for provisioning of scalable and measurable computed resources as needed by operational forecasting systems. It offers solutions for more flexible and adaptable computing architecture, for developing and running models, for managing and disseminating data to finally deploy services and applications. The review discussed on the key characteristic of cloud computing related on on-demand self-service, network access, resource pooling, elasticity and measured services. Additionally, it provides an overview of existing service models and deployments methods (e.g., private cloud, public cloud, community cloud, and hybrid cloud). A series of examples from the weather and ocean community is also briefly outlined, demonstrating how specific tasks can be mapped on specific cloud patterns and which methods are needed to be implemented depending on the specific adopted service model.Source: STATE OF THE PLANET, vol. 5 (issue 24)
DOI: 10.5194/sp-2024-37
Metrics:
Ciliberti S., Coro G.
Cloud computing offers an opportunity to innovate traditional methods for provisioning of scalable and measurable computed resources as needed by operational forecasting systems. It offers solutions for more flexible and adaptable computing architecture, for developing and running models, for managing and disseminating data to finally deploy services and applications. The review discussed on the key characteristic of cloud computing related on on-demand self-service, network access, resource pooling, elasticity and measured services. Additionally, it provides an overview of existing service models and deployments methods (e.g., private cloud, public cloud, community cloud, and hybrid cloud). A series of examples from the weather and ocean community is also briefly outlined, demonstrating how specific tasks can be mapped on specific cloud patterns and which methods are needed to be implemented depending on the specific adopted service model.Source: STATE OF THE PLANET, vol. 5 (issue 24)
DOI: 10.5194/sp-2024-37
Metrics:
See at:
doi.org | CNR IRIS | sp.copernicus.org | CNR IRIS | CNR IRIS | Copernicus Publications
2025
Journal article
Open Access
An open data collection of 3D tool and equipment models for neonatology
Bardelli S., Coro G., Scaramuzzo R. T., Ciantelli M., Cuttano A.
Virtual Simulation (VS) offers an elegant and effective solution to the current need for innovation in medical education, thanks to the possibility of creating low-cost, realistic training environments for repetitive practice without compromising patient safety. However, this training methodology is only adopted in some healthcare settings often because of the absence of free digital libraries of clinical assets and tools. The present technical note describes a data collection of 3D models representing crucial tools and equipment used in maternal and newborn care training. We used free-to-use photogrammetry and structure-from-motion software and computational platform for 3D object reconstruction to digitalize the physical clinical instruments typically used during maternal and newborn care. In particular, we acquired photographs of 34 physical objects and reconstructed them as 3D models. Additionally, we created a complete, navigable virtual training room containing the 3D models. Eventually, we published the 3D models and the virtual training room as an open-access data collection on Sketchfab (a free-to-use online digital platform for 3D model publication), from which all models can be freely downloaded and inspected through Web browsers, mobile applications, and Virtual and Augmented Reality devices. Our data collection and repeatable and cost-effective methodology open new opportunities to use VS for training through simulation in healthcare.Source: RESULTS IN ENGINEERING, vol. 25
DOI: 10.1016/j.rineng.2025.104236
Metrics:
Bardelli S., Coro G., Scaramuzzo R. T., Ciantelli M., Cuttano A.
Virtual Simulation (VS) offers an elegant and effective solution to the current need for innovation in medical education, thanks to the possibility of creating low-cost, realistic training environments for repetitive practice without compromising patient safety. However, this training methodology is only adopted in some healthcare settings often because of the absence of free digital libraries of clinical assets and tools. The present technical note describes a data collection of 3D models representing crucial tools and equipment used in maternal and newborn care training. We used free-to-use photogrammetry and structure-from-motion software and computational platform for 3D object reconstruction to digitalize the physical clinical instruments typically used during maternal and newborn care. In particular, we acquired photographs of 34 physical objects and reconstructed them as 3D models. Additionally, we created a complete, navigable virtual training room containing the 3D models. Eventually, we published the 3D models and the virtual training room as an open-access data collection on Sketchfab (a free-to-use online digital platform for 3D model publication), from which all models can be freely downloaded and inspected through Web browsers, mobile applications, and Virtual and Augmented Reality devices. Our data collection and repeatable and cost-effective methodology open new opportunities to use VS for training through simulation in healthcare.Source: RESULTS IN ENGINEERING, vol. 25
DOI: 10.1016/j.rineng.2025.104236
Metrics:
See at:
CNR IRIS | www.sciencedirect.com | CNR IRIS
2025
Journal article
Open Access
A FAIR and open geographic data collection for the Massaciuccoli Lake basin wetland in Italy
Vannini G. L., Bove P., Coro G.
The creation of a catalogue of geodata harmonised over time and space is essential for describing the status of ecosystem services in wetlands. In the present work, a specific methodology has been developed for the collection and generation of spatially and temporally harmonized geographic data to describe essential ecological and socio-economic charactesristics of the Massaciuccoli Lake basin (Tuscany, Italy), while providing a re-usable methodology for other areas. We developed a methodology, which we called ‘Geodata Layers Harmonization Methodology’ (GLHM), divided into four main phases: Geodata Census (GC), Geodata Selection (GS), Geodata Alignment (GA), and Geodata Publication (GP). The first phase, GC, involved a census of geodata made available online by public institutions, prioritizing those most relevant for describing ecosystem services, such as climatic, agro-environmental, pedo-geological, and biodiversity variables, with a preference for detailed data at the local level. The metadata of the collected geodata were organized into a structured tabular format. In the GS phase, geodata were selected based on a spatial resolution compatible with regional-scale ecological models (maximum 0.0005° ≈ 50 m), and a temporal coverage that could represent from remote past to far future scenarios. Geodata with partial spatial coverage or unsuitable for ecological models were excluded. Additionally, we evaluated the compliance of the geodata published on the websites of public institutions with the Findable-Accessible-Interoperable-Reusable (FAIR) principles through a newly developed scoring system. Based on this score, we selected only the data that exceeded a minimum FAIRness threshold. In the GA phase, the selected geodata were aligned semantically (i.e., by variable meaning), temporally, and spatially. Each geodata was georeferenced using the WGS84/EPSG:4326 reference system and clipped to the boundaries of the Massaciuccoli Lake basin. Raster data were resampled to achieve a uniform spatial resolution of 0.0005°. In the last phase, GP, the aligned geodata were published on public access repositories and services: The entire collection was organized as a QGIS project with legends and a metadata table associated. An Atlas was also produced, in PDF format, which visually represented the data and metadata. The geodata and their corresponding legends were exposed through Web Map Service (WMS) and Web feature Service (WFS) standards on a GeoServer instance and catalogued in a GeoNetwork instance, compliant with the ISO19139 standard and the INSPIRE European Directive. The collection contains 148 geo-datasets, representing 75 climatic, agro-environmental, pedo-geological, morphological, ecological, biological, and socio-economic information distributed across five temporal reference time frames: a remote past (1950–1980), a near past (1981–2015), the present (2016–2024), a near future (2025–2050), and a far future (2051–2100). Future projections are available under the Representative Concentration Pathways (RPC) 2.6, 4.5, and 8.5 to simulate low, medium, and high greenhouse gas concentration scenarios respectively. The present geodata collection is particularly useful for wetland monitoring, management and planning. It can easily be integrated with ecological models and predictive studies to analyse the effects of climate change and anthropogenic pressures on wetlands. The GLHM methodology is applicable to other ecological contexts to create standardised structured frameworks for evaluating the status of the biodiversity and the ecosystem services and the interplay between anthropic pressures and the ecosystem response.Source: DATA IN BRIEF, vol. 59
DOI: 10.1016/j.dib.2025.111303
Project(s): EcoScope
Metrics:
Vannini G. L., Bove P., Coro G.
The creation of a catalogue of geodata harmonised over time and space is essential for describing the status of ecosystem services in wetlands. In the present work, a specific methodology has been developed for the collection and generation of spatially and temporally harmonized geographic data to describe essential ecological and socio-economic charactesristics of the Massaciuccoli Lake basin (Tuscany, Italy), while providing a re-usable methodology for other areas. We developed a methodology, which we called ‘Geodata Layers Harmonization Methodology’ (GLHM), divided into four main phases: Geodata Census (GC), Geodata Selection (GS), Geodata Alignment (GA), and Geodata Publication (GP). The first phase, GC, involved a census of geodata made available online by public institutions, prioritizing those most relevant for describing ecosystem services, such as climatic, agro-environmental, pedo-geological, and biodiversity variables, with a preference for detailed data at the local level. The metadata of the collected geodata were organized into a structured tabular format. In the GS phase, geodata were selected based on a spatial resolution compatible with regional-scale ecological models (maximum 0.0005° ≈ 50 m), and a temporal coverage that could represent from remote past to far future scenarios. Geodata with partial spatial coverage or unsuitable for ecological models were excluded. Additionally, we evaluated the compliance of the geodata published on the websites of public institutions with the Findable-Accessible-Interoperable-Reusable (FAIR) principles through a newly developed scoring system. Based on this score, we selected only the data that exceeded a minimum FAIRness threshold. In the GA phase, the selected geodata were aligned semantically (i.e., by variable meaning), temporally, and spatially. Each geodata was georeferenced using the WGS84/EPSG:4326 reference system and clipped to the boundaries of the Massaciuccoli Lake basin. Raster data were resampled to achieve a uniform spatial resolution of 0.0005°. In the last phase, GP, the aligned geodata were published on public access repositories and services: The entire collection was organized as a QGIS project with legends and a metadata table associated. An Atlas was also produced, in PDF format, which visually represented the data and metadata. The geodata and their corresponding legends were exposed through Web Map Service (WMS) and Web feature Service (WFS) standards on a GeoServer instance and catalogued in a GeoNetwork instance, compliant with the ISO19139 standard and the INSPIRE European Directive. The collection contains 148 geo-datasets, representing 75 climatic, agro-environmental, pedo-geological, morphological, ecological, biological, and socio-economic information distributed across five temporal reference time frames: a remote past (1950–1980), a near past (1981–2015), the present (2016–2024), a near future (2025–2050), and a far future (2051–2100). Future projections are available under the Representative Concentration Pathways (RPC) 2.6, 4.5, and 8.5 to simulate low, medium, and high greenhouse gas concentration scenarios respectively. The present geodata collection is particularly useful for wetland monitoring, management and planning. It can easily be integrated with ecological models and predictive studies to analyse the effects of climate change and anthropogenic pressures on wetlands. The GLHM methodology is applicable to other ecological contexts to create standardised structured frameworks for evaluating the status of the biodiversity and the ecosystem services and the interplay between anthropic pressures and the ecosystem response.Source: DATA IN BRIEF, vol. 59
DOI: 10.1016/j.dib.2025.111303
Project(s): EcoScope
Metrics:
See at:
CNR IRIS | www.sciencedirect.com | CNR IRIS
2025
Journal article
Open Access
Phoné: an Initiative to develop a dataset for the automatic recognition of spoken Italian
Coro G., Cutugno F., Schettino L., Tanda E., Vietti A., Vitale V. N.
Large Language Models (LLM) have revolutionised natural language processing and its applications. However, high-performance LLMs require copious data and computing resources for their development and are rarely public. This also concerns Large Acoustic Models (LAM) for processing spoken language. The Phoné initiative seeks to build an open Italian speech dataset to advance Automatic Speech Recognition (ASR) systems and support public research. Spearheaded by institutions in Naples, Pisa, and Bolzano, the project gathers diverse Italian audio sources and applies advanced ASR architectures, including supervised and self-supervised models. This paper details Phoné’s dataset creation, ASR model evaluation, and ethical considerations, aiming to democratise access to Italian-language resources and foster innovation in ASR technologies.Source: ORAL ARCHIVES JOURNAL, vol. 1
DOI: 10.36253/oar-3340
Metrics:
Coro G., Cutugno F., Schettino L., Tanda E., Vietti A., Vitale V. N.
Large Language Models (LLM) have revolutionised natural language processing and its applications. However, high-performance LLMs require copious data and computing resources for their development and are rarely public. This also concerns Large Acoustic Models (LAM) for processing spoken language. The Phoné initiative seeks to build an open Italian speech dataset to advance Automatic Speech Recognition (ASR) systems and support public research. Spearheaded by institutions in Naples, Pisa, and Bolzano, the project gathers diverse Italian audio sources and applies advanced ASR architectures, including supervised and self-supervised models. This paper details Phoné’s dataset creation, ASR model evaluation, and ethical considerations, aiming to democratise access to Italian-language resources and foster innovation in ASR technologies.Source: ORAL ARCHIVES JOURNAL, vol. 1
DOI: 10.36253/oar-3340
Metrics:
See at:
CNR IRIS | riviste.fupress.net | CNR IRIS
2025
Journal article
Open Access
Managing fisheries and ecosystems: current good practices and the EcoScope project experience
Daskalov G., De La Puente S., Scotti M., Klayn S., Briguglio M., Coro G., Gal G., Heymans J. J., Rodriguez-Perez A., Steenbeek J.
Ecosystem Based Management (EBM) is a comprehensive way of managing fisheries and marine resources. As such, it needs a large and complex suite of concepts and tools to address a variety of problems ranging from climate change, through various forms of water pollution, to trophic interactions and social-economic sustainability. Industry, scientists, managers, and policy makers involved in the fisheries sector are the main actors in EBM. EBM objectives based on policy needs, legal requirements, and ecosystem considerations may target specific fish stocks, or encompass several ecosystem components aiming for balanced fisheries, but they need to address the trade-offs between maximizing economic gains versus sustainable fisheries and healthy ecosystems. Fishing at Maximum Sustainable Yield (MSY), setting ecosystem reference points, discards ban, avoiding bycatch of protected species, habitat protection, accounting for the effects of climate change, achieving good environmental status, setting effective marine protected areas, and considering ecosystem effects from marine spatial planning, are all examples of EBM objectives. The EcoScope project aimed to address ecosystem degradation, anthropogenic impacts, and unsustainable fisheries by developing an efficient, holistic, ecosystem-based approach to sustainable fisheries management that can easily be used by policy makers and advisory bodies. The EcoScope consortium reflects an interdisciplinary advisory team of biologists, modelers, economists, and social scientists. It performed comprehensive reviews of data, data gaps, and various tools (models, indicators, management evaluation procedures). An online platform, toolbox, academy, and a mobile application are end products delivered and maintained by EcoScope to facilitate knowledge sharing, communication, and education. The EcoScope project has built modules ready to be used in the implementation of EBM, but a more direct approach by the responsible organizations, such as ICES, FAO, GFCM and the EC, is needed to set explicit and formal research and managerial frameworks for implementing and coordinating the EBM activities.Source: FRONTIERS IN MARINE SCIENCE, vol. 12
DOI: 10.3389/fmars.2025.1640487
Project(s): EcoScope
Metrics:
Daskalov G., De La Puente S., Scotti M., Klayn S., Briguglio M., Coro G., Gal G., Heymans J. J., Rodriguez-Perez A., Steenbeek J.
Ecosystem Based Management (EBM) is a comprehensive way of managing fisheries and marine resources. As such, it needs a large and complex suite of concepts and tools to address a variety of problems ranging from climate change, through various forms of water pollution, to trophic interactions and social-economic sustainability. Industry, scientists, managers, and policy makers involved in the fisheries sector are the main actors in EBM. EBM objectives based on policy needs, legal requirements, and ecosystem considerations may target specific fish stocks, or encompass several ecosystem components aiming for balanced fisheries, but they need to address the trade-offs between maximizing economic gains versus sustainable fisheries and healthy ecosystems. Fishing at Maximum Sustainable Yield (MSY), setting ecosystem reference points, discards ban, avoiding bycatch of protected species, habitat protection, accounting for the effects of climate change, achieving good environmental status, setting effective marine protected areas, and considering ecosystem effects from marine spatial planning, are all examples of EBM objectives. The EcoScope project aimed to address ecosystem degradation, anthropogenic impacts, and unsustainable fisheries by developing an efficient, holistic, ecosystem-based approach to sustainable fisheries management that can easily be used by policy makers and advisory bodies. The EcoScope consortium reflects an interdisciplinary advisory team of biologists, modelers, economists, and social scientists. It performed comprehensive reviews of data, data gaps, and various tools (models, indicators, management evaluation procedures). An online platform, toolbox, academy, and a mobile application are end products delivered and maintained by EcoScope to facilitate knowledge sharing, communication, and education. The EcoScope project has built modules ready to be used in the implementation of EBM, but a more direct approach by the responsible organizations, such as ICES, FAO, GFCM and the EC, is needed to set explicit and formal research and managerial frameworks for implementing and coordinating the EBM activities.Source: FRONTIERS IN MARINE SCIENCE, vol. 12
DOI: 10.3389/fmars.2025.1640487
Project(s): EcoScope
Metrics:
See at:
CNR IRIS | www.frontiersin.org | CNR IRIS
2025
Dataset
Open Access
Massaciuccoli Lake basin - Cartographic Comparison of Spatial Analysis Methodologies for Ecosystem Risk Assessment
Vannini G. L., Coro G.
This geographic collection presents a cartographic comparison of three spatial analysis methodologies—Overlap, Multi K-means clustering, and Multi K-means applied to Variational Autoencoder (VAE) outputs—aimed at identifying the Risk of forest and grassland habitat degradation and the associated biodiversity loss.DOI: 10.5281/zenodo.17160216
Project(s): Integrated Environmental Research Infrastructures System
Metrics:
Vannini G. L., Coro G.
This geographic collection presents a cartographic comparison of three spatial analysis methodologies—Overlap, Multi K-means clustering, and Multi K-means applied to Variational Autoencoder (VAE) outputs—aimed at identifying the Risk of forest and grassland habitat degradation and the associated biodiversity loss.DOI: 10.5281/zenodo.17160216
Project(s): Integrated Environmental Research Infrastructures System
Metrics:
2025
Other
Restricted
InfraScience research activity report 2024
Angioni S., Artini M., Assante M., Atzori C., Baglioni M., Bardi A., Bosio C., Bove P., Calanducci A., Candela L., Casini G., Castelli D., Cirillo R., Coro G., De Bonis M., Debole F., Dell'Amico A., Frosini L., Ibrahim Ahmed, La Bruzzo S., Lelii L., Manghi P., Mangiacrapa F., Mangione D., Mannocci A., Molinaro E., Oliviero A., Pagano P., Panichi G., Teresa M. T., Pavone G., Peccerillo B., Piccioli T., Procaccini M., Straccia U., Vannini G. L., Versienti L.
InfraScience is a research group within the Institute of Information Science and Technologies (ISTI) of the National Research Council of Italy (CNR), based in Pisa. This activity report outlines the group's research achievements and initiatives throughout 2024. InfraScience focused its efforts on key challenges in the areas of Data Infrastructures, e-Science, and Intelligent Systems, maintaining a strong synergy between research and development and a firm commitment to open science principles. In 2024, the group played a leading role in the development and evolution of two major Open Science infrastructures: D4Science and OpenAIRE. InfraScience researchers contributed significantly to the scientific community through the publication of peer-reviewed papers, active participation in EU-funded research projects, organization of international conferences and training activities, and engagement in various working groups and task forces. This report highlights these contributions and underscores the group's ongoing dedication to advancing open, collaborative, and impactful science.DOI: 10.32079/isti-ar-2025/001
Metrics:
Angioni S., Artini M., Assante M., Atzori C., Baglioni M., Bardi A., Bosio C., Bove P., Calanducci A., Candela L., Casini G., Castelli D., Cirillo R., Coro G., De Bonis M., Debole F., Dell'Amico A., Frosini L., Ibrahim Ahmed, La Bruzzo S., Lelii L., Manghi P., Mangiacrapa F., Mangione D., Mannocci A., Molinaro E., Oliviero A., Pagano P., Panichi G., Teresa M. T., Pavone G., Peccerillo B., Piccioli T., Procaccini M., Straccia U., Vannini G. L., Versienti L.
InfraScience is a research group within the Institute of Information Science and Technologies (ISTI) of the National Research Council of Italy (CNR), based in Pisa. This activity report outlines the group's research achievements and initiatives throughout 2024. InfraScience focused its efforts on key challenges in the areas of Data Infrastructures, e-Science, and Intelligent Systems, maintaining a strong synergy between research and development and a firm commitment to open science principles. In 2024, the group played a leading role in the development and evolution of two major Open Science infrastructures: D4Science and OpenAIRE. InfraScience researchers contributed significantly to the scientific community through the publication of peer-reviewed papers, active participation in EU-funded research projects, organization of international conferences and training activities, and engagement in various working groups and task forces. This report highlights these contributions and underscores the group's ongoing dedication to advancing open, collaborative, and impactful science.DOI: 10.32079/isti-ar-2025/001
Metrics:
2025
Other
Open Access
Massaciuccoli Lake Basin: risk hotspots of forest and grassland habitat degradation and biodiversity loss
Vannini G. L., Coro G.
This technical report presents a cartographic representation of ecosystem risk hotspots related to forest and grassland habitat degradation and biodiversity loss within the wetland area of the Massaciuccoli Lake basin (Tuscany, Italy). The map also highlights the main environmental and anthropogenic stressors contributing to these risks.DOI: 10.5281/zenodo.17369644
DOI: 10.5281/zenodo.17369645
DOI: 10.5281/zenodo.17414332
Project(s): Integrated Environmental Research Infrastructures System
Metrics:
Vannini G. L., Coro G.
This technical report presents a cartographic representation of ecosystem risk hotspots related to forest and grassland habitat degradation and biodiversity loss within the wetland area of the Massaciuccoli Lake basin (Tuscany, Italy). The map also highlights the main environmental and anthropogenic stressors contributing to these risks.DOI: 10.5281/zenodo.17369644
DOI: 10.5281/zenodo.17369645
DOI: 10.5281/zenodo.17414332
Project(s): Integrated Environmental Research Infrastructures System
Metrics:
See at:
IRIS Cnr | IRIS Cnr | IRIS Cnr | IRIS Cnr | CNR IRIS | IRIS Cnr | ZENODO | ZENODO | ZENODO | CNR IRIS
2025
Journal article
Open Access
Assessing marine ecosystem risks through unsupervised methods
Laura Pavirani, Pasquale Bove, Gianpaolo Coro
Marine ecosystems are facing significant challenges from intensified fishing, pollution, climate change, and biodiversity loss. Ecosystem risk assessments are vital for informing effective policies and management decisions. Traditional approaches, such as Ecosystem Models (EMs) and Marine Spatial Planning (MSP), often rely on expert knowledge, which introduces subjective assumptions. This study evaluates six unsupervised methods — four clustering algorithms (Multi K-means, Fuzzy C-means, X-means, and DBSCAN) and two machine-learning models (an Artificial Neural Network, ANN, and a Variational Autoencoder, VAE) — to assess marine ecosystem risk in the Mediterranean Sea automatically, using open-access data from 2017 to 2021. Each method generated five annual high-risk maps based on ecosystem variables, including fishing effort, species richness, depth, coastal proximity, oxygen levels, net primary production, and thermohaline circulation intensity. Our quantitative analysis of 30 generated maps revealed pairwise similarities ranging from 72.2% to 95.9%, with Cohen’s Kappa scores between 0.46 (moderate) and 0.91 (almost perfect). All methods consistently identified high-risk hotspots in the Eastern and Western Mediterranean, the Tyrrhenian Sea, the Adriatic Sea, the Strait of Sicily, and the Aegean Sea. However, we also found discrepancies due to the different tendencies of the models to produce broader (precautionary) or more focused (conservative) risk assessments. Assessments by DBSCAN, ANN, and VAE were similar (∼90%) and broader, whereas X-means was more conservative. Multi K-means and Fuzzy C-means exhibited similar (∼92%) and more balanced results. These findings provide a data-driven foundation and practical guidance for developing Bayesian EMs and MSP with reduced reliance on subjective assessments.Source: ECOLOGICAL INFORMATICS, vol. 90 (issue December), pp. 1-16
DOI: 10.1016/j.ecoinf.2025.103334
Project(s): EcoScope
Metrics:
Laura Pavirani, Pasquale Bove, Gianpaolo Coro
Marine ecosystems are facing significant challenges from intensified fishing, pollution, climate change, and biodiversity loss. Ecosystem risk assessments are vital for informing effective policies and management decisions. Traditional approaches, such as Ecosystem Models (EMs) and Marine Spatial Planning (MSP), often rely on expert knowledge, which introduces subjective assumptions. This study evaluates six unsupervised methods — four clustering algorithms (Multi K-means, Fuzzy C-means, X-means, and DBSCAN) and two machine-learning models (an Artificial Neural Network, ANN, and a Variational Autoencoder, VAE) — to assess marine ecosystem risk in the Mediterranean Sea automatically, using open-access data from 2017 to 2021. Each method generated five annual high-risk maps based on ecosystem variables, including fishing effort, species richness, depth, coastal proximity, oxygen levels, net primary production, and thermohaline circulation intensity. Our quantitative analysis of 30 generated maps revealed pairwise similarities ranging from 72.2% to 95.9%, with Cohen’s Kappa scores between 0.46 (moderate) and 0.91 (almost perfect). All methods consistently identified high-risk hotspots in the Eastern and Western Mediterranean, the Tyrrhenian Sea, the Adriatic Sea, the Strait of Sicily, and the Aegean Sea. However, we also found discrepancies due to the different tendencies of the models to produce broader (precautionary) or more focused (conservative) risk assessments. Assessments by DBSCAN, ANN, and VAE were similar (∼90%) and broader, whereas X-means was more conservative. Multi K-means and Fuzzy C-means exhibited similar (∼92%) and more balanced results. These findings provide a data-driven foundation and practical guidance for developing Bayesian EMs and MSP with reduced reliance on subjective assessments.Source: ECOLOGICAL INFORMATICS, vol. 90 (issue December), pp. 1-16
DOI: 10.1016/j.ecoinf.2025.103334
Project(s): EcoScope
Metrics:
See at:
CNR IRIS | www.sciencedirect.com | CNR IRIS | CNR IRIS
2025
Conference article
Restricted
Ecosystem risk Aassessment through stressor concurrency identification: a comparative analysis
Laura Pavirani, Pasquale Bove, Gianpaolo Coro
In Marine Science, ecosystem risk assessment is a process that integrates data to estimate the potential impact of harmful and fragile forces (stressors) on the ecosystem. Nowadays, the management of marine ecosystems is increasingly complex due to multiple stressors, including human activities and climate change, and requires robust tools to address challenges effectively. We present big data-driven methods that enable a rapid, simultaneous analysis of multiple stressors using unsupervised learning techniques and statistical analysis to produce prior ecosystem risk assessments. We apply four cluster analysis methods based on Multi K-means, Fuzzy C-means, X-means, and DBSCAN, to identify stressor concurrency areas in Mediterranean Sea data from 2017 to 2021. These data include stressor variables related to environmental, oceanographic, fishing, and biodiversity factors. The methods assess ecosystem risk by detecting high stressor concurrency conditions. Finally, they produce maps that highlight potential high-risk regions. We compare the results of the four methods to examine the similarities and differences in their abilities to detect high-risk areas. From the Mediterranean data, all methods jointly indicate known high-risk areas but differ in the extent of the identified areas. Our comparative analysis highlights the importance of selecting the most appropriate clustering technique based on the balance between precautionary (highlighting broader areas) and conservative (highlighting smaller areas) perspectives. The results provide information that should be used in ecosystem models and marine spatial planning to improve the accuracy and objectivity of ecosystem risk assessment and management strategies.DOI: 10.1109/oceans58557.2025.11104682
Project(s): EcoScope, ITINERIS PNRR Italian project
Metrics:
Laura Pavirani, Pasquale Bove, Gianpaolo Coro
In Marine Science, ecosystem risk assessment is a process that integrates data to estimate the potential impact of harmful and fragile forces (stressors) on the ecosystem. Nowadays, the management of marine ecosystems is increasingly complex due to multiple stressors, including human activities and climate change, and requires robust tools to address challenges effectively. We present big data-driven methods that enable a rapid, simultaneous analysis of multiple stressors using unsupervised learning techniques and statistical analysis to produce prior ecosystem risk assessments. We apply four cluster analysis methods based on Multi K-means, Fuzzy C-means, X-means, and DBSCAN, to identify stressor concurrency areas in Mediterranean Sea data from 2017 to 2021. These data include stressor variables related to environmental, oceanographic, fishing, and biodiversity factors. The methods assess ecosystem risk by detecting high stressor concurrency conditions. Finally, they produce maps that highlight potential high-risk regions. We compare the results of the four methods to examine the similarities and differences in their abilities to detect high-risk areas. From the Mediterranean data, all methods jointly indicate known high-risk areas but differ in the extent of the identified areas. Our comparative analysis highlights the importance of selecting the most appropriate clustering technique based on the balance between precautionary (highlighting broader areas) and conservative (highlighting smaller areas) perspectives. The results provide information that should be used in ecosystem models and marine spatial planning to improve the accuracy and objectivity of ecosystem risk assessment and management strategies.DOI: 10.1109/oceans58557.2025.11104682
Project(s): EcoScope
, ITINERIS PNRR Italian projectMetrics:
See at:
CNR IRIS | ieeexplore.ieee.org | CNR IRIS | CNR IRIS
2024
Journal article
Open Access
Exploring emergent syllables in end-to-end automatic speech recognizers through model explainability technique
Vitale Vincenzo Norman, Cutugno Francesco, Origlia Antonio, Coro Gianpaolo
Automatic speech recognition systems based on end-to-end models (E2E-ASRs) can achieve comparable performance to conventional ASR systems while reproducing all their essential parts automatically, from speech units to the language model. However, they hide the underlying perceptual processes modelled, if any, and they have lower adaptability to multiple application contexts, and, furthermore, they require powerful hardware and an extensive amount of training data. Model-explainability techniques can explore the internal dynamics of these ASR systems and possibly understand and explain the processes conducting to their decisions and outputs. Understanding these processes can help enhance ASR performance and reduce the required training data and hardware significantly. In this paper, we probe the internal dynamics of three E2E-ASRs pre-trained for English by building an acoustic-syllable boundary detector for Italian and Spanish based on the E2E-ASRs’ internal encoding layer outputs. We demonstrate that the shallower E2E-ASR layers spontaneously form a rhythmic component correlated with prominent syllables, central in human speech processing. This finding highlights a parallel between the analysed E2E-ASRs and human speech recognition. Our results contribute to the body of knowledge by providing a human-explainable insight into behaviours encoded in popular E2E-ASR systems.Source: NEURAL COMPUTING & APPLICATIONS, vol. 36, pp. 6875-6901
DOI: 10.1007/s00521-024-09435-1
Metrics:
Vitale Vincenzo Norman, Cutugno Francesco, Origlia Antonio, Coro Gianpaolo
Automatic speech recognition systems based on end-to-end models (E2E-ASRs) can achieve comparable performance to conventional ASR systems while reproducing all their essential parts automatically, from speech units to the language model. However, they hide the underlying perceptual processes modelled, if any, and they have lower adaptability to multiple application contexts, and, furthermore, they require powerful hardware and an extensive amount of training data. Model-explainability techniques can explore the internal dynamics of these ASR systems and possibly understand and explain the processes conducting to their decisions and outputs. Understanding these processes can help enhance ASR performance and reduce the required training data and hardware significantly. In this paper, we probe the internal dynamics of three E2E-ASRs pre-trained for English by building an acoustic-syllable boundary detector for Italian and Spanish based on the E2E-ASRs’ internal encoding layer outputs. We demonstrate that the shallower E2E-ASR layers spontaneously form a rhythmic component correlated with prominent syllables, central in human speech processing. This finding highlights a parallel between the analysed E2E-ASRs and human speech recognition. Our results contribute to the body of knowledge by providing a human-explainable insight into behaviours encoded in popular E2E-ASR systems.Source: NEURAL COMPUTING & APPLICATIONS, vol. 36, pp. 6875-6901
DOI: 10.1007/s00521-024-09435-1
Metrics:
See at:
Neural Computing and Applications | FEDOA - IRIS Università degli Studi Napoli Federico II | CNR IRIS | CNR IRIS
2024
Journal article
Open Access
An open science automatic workflow for multi-model species distribution estimation
Coro Gianpaolo, Sana Lorenzo, Bove Pasquale
Integrated Environmental Assessment systems and ecosystem models study the links between anthropogenic and climatic pressures on marine ecosystems and help understand how to manage the effects of the unsustainable exploitation of ocean resources. However, these models have long implementation times, data and model interoperability issues and require heterogeneous competencies. Therefore, they would benefit from simplification, automatisation, and enhanced integrability of the underlying models. Artificial Intelligence can help overcome several limitations by speeding up the modelling of crucial functional parts, e.g. estimating the environmental conditions fostering a species’ persistence and proliferation in an area (the species’ ecological niche) and, consequently, its geographical distribution. This paper presents a full-automatic workflow to estimate species’ distributions through statistical and machine learning models. It embeds four ecological niche models with complementary approaches, i.e. Artificial Neural Networks, Maximum Entropy, Support Vector Machines, and AquaMaps. It automatically estimates the optimal model parametrisations and decision thresholds to distinguish between suitable- and unsuitable-habitat locations and combines the models within one ensemble model. Finally, it combines several ensemble models to produce a species richness map (biodiversity index). The software is open-source, Open Science compliant, and available as a Web Processing Service-standardised cloud computing service that enhances efficiency, integrability, cross-domain reusability, and experimental reproduction and repetition. We first assess workflow stability and sensitivity and then demonstrate effectiveness by producing a biodiversity index for the Mediterranean based on $$\sim $$1500 species data. Moreover, we predict the spread of the invasive Siganus rivulatus in the Mediterranean and its current and future overlap with the native Sarpa salpa under different climate change scenarios.Source: INTERNATIONAL JOURNAL OF DATA SCIENCE AND ANALYTICS
DOI: 10.1007/s41060-024-00517-w
Project(s): EcoScope, ITINERIS PNRR Italian project
Metrics:
Coro Gianpaolo, Sana Lorenzo, Bove Pasquale
Integrated Environmental Assessment systems and ecosystem models study the links between anthropogenic and climatic pressures on marine ecosystems and help understand how to manage the effects of the unsustainable exploitation of ocean resources. However, these models have long implementation times, data and model interoperability issues and require heterogeneous competencies. Therefore, they would benefit from simplification, automatisation, and enhanced integrability of the underlying models. Artificial Intelligence can help overcome several limitations by speeding up the modelling of crucial functional parts, e.g. estimating the environmental conditions fostering a species’ persistence and proliferation in an area (the species’ ecological niche) and, consequently, its geographical distribution. This paper presents a full-automatic workflow to estimate species’ distributions through statistical and machine learning models. It embeds four ecological niche models with complementary approaches, i.e. Artificial Neural Networks, Maximum Entropy, Support Vector Machines, and AquaMaps. It automatically estimates the optimal model parametrisations and decision thresholds to distinguish between suitable- and unsuitable-habitat locations and combines the models within one ensemble model. Finally, it combines several ensemble models to produce a species richness map (biodiversity index). The software is open-source, Open Science compliant, and available as a Web Processing Service-standardised cloud computing service that enhances efficiency, integrability, cross-domain reusability, and experimental reproduction and repetition. We first assess workflow stability and sensitivity and then demonstrate effectiveness by producing a biodiversity index for the Mediterranean based on $$\sim $$1500 species data. Moreover, we predict the spread of the invasive Siganus rivulatus in the Mediterranean and its current and future overlap with the native Sarpa salpa under different climate change scenarios.Source: INTERNATIONAL JOURNAL OF DATA SCIENCE AND ANALYTICS
DOI: 10.1007/s41060-024-00517-w
Project(s): EcoScope
, ITINERIS PNRR Italian projectMetrics:
See at:
International Journal of Data Science and Analytics | CNR IRIS | CNR IRIS
2024
Dataset
Open Access
Interactive geographic catalog of environmental, geomorphologic, and socio-economic variables of the Massaciuccoli Lake basin in Tuscany, Italy
Vannini G. L., Coro G., Panichi G.
The catalogue is the result of a process of evaluation, selection and harmonization of environmental, geomorphological and socio-economic data. The data were acquired following the evaluation of FAIR and Open Access principles, from varied sources with heterogeneous spatio-temporal resolutions. The data were then transformed into spatiotemporally aligned datasets and described in a standardised form. The metadata are described in the ISO 19139 standard, in accordance with the INSPIRE directives. The visualisation application integrated in GeoNetwork allows a simple overlay of the data (on first viewing, 'layerExtentZoom' must be selected from the options of the first selected layer). All data are open access.Project(s): Integrated Environmental Research Infrastructures System
Vannini G. L., Coro G., Panichi G.
The catalogue is the result of a process of evaluation, selection and harmonization of environmental, geomorphological and socio-economic data. The data were acquired following the evaluation of FAIR and Open Access principles, from varied sources with heterogeneous spatio-temporal resolutions. The data were then transformed into spatiotemporally aligned datasets and described in a standardised form. The metadata are described in the ISO 19139 standard, in accordance with the INSPIRE directives. The visualisation application integrated in GeoNetwork allows a simple overlay of the data (on first viewing, 'layerExtentZoom' must be selected from the options of the first selected layer). All data are open access.Project(s): Integrated Environmental Research Infrastructures System
See at:
CNR IRIS | services.d4science.org | CNR IRIS
2024
Journal article
Open Access
A semantic knowledge graph of European mountain value chains
Bartalesi Lenzi V., Coro G., Lenzi E., Pratelli N., Pagano P., Moretti M., Brunori G.
The United Nations forecast a significant shift in global population distribution by 2050, with rural populations projected to decline. This decline will particularly challenge mountain areas' cultural heritage, well-being, and economic sustainability. Understanding the economic, environmental, and societal effects of rural population decline is particularly important in Europe, where mountainous regions are vital for supplying goods. The present paper describes a geospatially explicit semantic knowledge graph containing information on 454 European mountain value chains. It is the first large-size, structured collection of information on mountain value chains. Our graph, structured through ontology-based semantic modelling, offers representations of the value chains in the form of narratives. The graph was constructed semi-automatically from unstructured data provided by mountain-area expert scholars. It is accessible through a public repository and explorable through interactive Story Maps and a semantic Web service. Through semantic queries, we demonstrate that the graph allows for exploring territorial complexities and discovering new knowledge on mountain areas' environmental, societal, territory, and economic aspects that could help stem depopulation.Source: SCIENTIFIC DATA, vol. 11
DOI: 10.1038/s41597-024-03760-9
Project(s): Mountain Valorization through Interconnectedness and Green Growth
Metrics:
Bartalesi Lenzi V., Coro G., Lenzi E., Pratelli N., Pagano P., Moretti M., Brunori G.
The United Nations forecast a significant shift in global population distribution by 2050, with rural populations projected to decline. This decline will particularly challenge mountain areas' cultural heritage, well-being, and economic sustainability. Understanding the economic, environmental, and societal effects of rural population decline is particularly important in Europe, where mountainous regions are vital for supplying goods. The present paper describes a geospatially explicit semantic knowledge graph containing information on 454 European mountain value chains. It is the first large-size, structured collection of information on mountain value chains. Our graph, structured through ontology-based semantic modelling, offers representations of the value chains in the form of narratives. The graph was constructed semi-automatically from unstructured data provided by mountain-area expert scholars. It is accessible through a public repository and explorable through interactive Story Maps and a semantic Web service. Through semantic queries, we demonstrate that the graph allows for exploring territorial complexities and discovering new knowledge on mountain areas' environmental, societal, territory, and economic aspects that could help stem depopulation.Source: SCIENTIFIC DATA, vol. 11
DOI: 10.1038/s41597-024-03760-9
Project(s): Mountain Valorization through Interconnectedness and Green Growth
Metrics:
See at:
Scientific Data | Scientific Data | IRIS Cnr | IRIS Cnr | Software Heritage | GitHub | CNR IRIS