34711 research outputs found
Sort by
Aquatic macroinvertebrates as indicators for hydropeaking : development and validation of specific multimetric indices
No full textHydropeaking imposes sub‑daily discharge fluctuations that, among others, alter hydraulics, wetted area, and benthic communities. Benthic habitat sampling and analyses are well suited to detect long-term changes in faunal composition resulting from cumulative hydropeaking effects. In this study, we developed and validated a hydropeaking‑sensitive multimetric index (MMI) using macroinvertebrate assemblages at 54 sites on 11 Swiss rivers spanning hydropeaked, residual‑flow, and (near‑)natural reaches. Standardized field‑screening (12 kick samples per site; 500 µm) was conducted in March–April 2022–2023; a subset of 31 sites (370 samples out of the total 648 samples) underwent laboratory identification for quality control. 42 environmental variables were compiled, and recent sub‑daily hydrology (15‑min, prior six months) was summarized, including eight hydropeaking-specific hydrological variables and a multimetric hydrological index. 25 suitable macroinvertebrate candidate community metrics were pre-selected and calculated for each study site located in reaches with a hydropeaking or a (near-) natural flow regime. These metrics were then evaluated based on their variability between laboratory and field-screening datasets, their indicative quality regarding hydropeaking sensitivity and natural variability, and how they were influenced by environmental variables such as morphological alterations and pollution.
The MMI developed from the obtained results integrates individual metrics that best represent the established aquatic macroinvertebrate community, while specifically reflecting hydropeaking-related hydrological and hydraulic impacts and minimizing the influence of natural variability and other environmental factors. This MMI allows for the comparison of different river reaches regardless of river type and can therefore be applied not only to assess hydropeaking impacts, but also to evaluate the effectiveness of hydrological and/or morphological mitigation or restoration measures through a pre- and post-implementation comparison of the benthic macroinvertebrate community – serving, in this sense, as an efficient monitoring tool
An exploratory analysis of LLM-based machine-translated audio description scripts in the French-German language pair
No full text(Semi)automated methods in audio description (AD) production, e.g. Large Language Model (LLM)-based machine translation (MT) of AD scripts can address the rising demand for ADs due to reinforced legal requirements (cf. Braun & Starr 2022). This is particularly true for multilingual countries such as Switzerland where AD scripts are currently being created ex novo in three national languages. Previous studies indicate that the use of MT could be an alternative production method in AD and recommend further investigation into post-editing effort (Fernández-Torné 2016; Fernández-Torné & Matamala 2016; Matamala & Ortiz-Boiz 2016; Vercauteren et al. 2021). Considering recent developments in artificial intelligence (AI) and the use of LLM-based MT, this becomes all the more important. In line with common practice in (human) MT quality assessment (cf. e.g. Bentivogli et al. 2018), we conducted a source-based direct assessment with professional audio describers to evaluate machine-translated AD segments in the French-German language pair (Lüthi 2024). AD segments had been translated with OpenAI’s GPT-4 model. AD professionals evaluated three different translation versions: 1) generated with textual input only; 2) generated with both textual and visual input, 3) the latter post-edited by ChatGPT-4. Evaluators were asked to assign adequacy, fluency and usefulness scores to every AD segment without knowing how each translation version had been produced. Additionally, we applied a qualitative approach and analyzed selected AD segments using the harmonized DQF-MQM error typology (cf. Lommel 2018). Our results suggest that LLM-based, human-assisted MT can be a viable alternative method to produce ready-for-voicing AD scripts.References:
Bentivogli, L., Cettolo, M., Marcello, F., & Federmann, C. (2018). Machine Translation Human Evaluation: An investigation of evaluation based on Post-Editing and its relation with Direct Assessment. Proceedings of the 15th International Workshop on Spoken Language Translation (IWSLT 2018), 62–69.
Braun, S., & Starr, K. (2022). Automating audio description. In C. Taylor & E. Perego (Eds.), The Routledge Handbook of Audio Description (pp. 391–406). Routledge. https://doi.org/10.4324/9781003003052-30
Fernández-Torné, A. (2016). Machine translation evaluation through post-editing measures in audio description. inTRAlinea 18. https://www.intralinea.org/archive/article/2200
Fernández-Torné, A., & Matamala, A. (2016). Machine translation in audio description? Comparing creation, translation and post-editing efforts. SKASE Journal of Translation and Interpretation, 9(1), 64–87.
Lommel, A. (2018). Metrics for Translation Quality Assessment: A Case for Standardising Error Typologies. In J. Moorkens, S. Castilho, F. Gaspari, & S. Doherty (Eds.), Translation Quality Assessment: From Principles to Practice (pp. 109–127). Springer. https://doi.org/10.1007/978-3-319-91241-7_6
Matamala, A., & Ortiz-Boix, C. (2016). Accessibility and multilingualism: an exploratory study on the machine translation of audio descriptions. TRANS, 20, 11–24. https://doi.org/10.24310/TRANS.2016.v0i20.2059
Lüthi, N. (2024). KI-gestützte (maschinelle) Übersetzung von Audiodeskription. Eine quantitative und qualitative Untersuchung der Sendung Passe-moi les jumelles. Graduate Papers in Applied Linguistics 26. Zurich University of Applied Sciences. https://doi.org/10.21256/zhaw-2823
Vercauteren, G., Reviers, N., & Steyaert, K. (2021). Evaluating the effectiveness of machine translation of audio description: the results of two pilot studies in the English-Dutch language pair. Revista Tradumàtica – Tecnologies de la Traducció, 19, 226–252. https://doi.org/10.5565/rev/tradumatica.28
Temperate agroforestry for tree carbon storage in Switzerland : 10 years of biophysical and social monitoring
No full textAgroforestry, the integration of woody structures in agricultural land, has high potential for climate protection and resilience, since trees are active carbon sinks. Yet, there is only limited empirical evidence on the actual performance of temperate agroforestry systems in this respect, nor on its acceptance by farmers. We monitored four silvoarable agroforestry systems in Switzerland (apple, sour cherry, poplar, wild cherry) over ten years and measured tree growth and carbon storage performances. We compared the measured data to outcomes of the Yield-SAFE model. We regularly interviewed farmers on their observations of their agroforestry systems. Individual growth of agroforestry trees varied between species and location, with differences between the smallest and largest tree ranging from 44 % to 97 %. Consequently, the carbon sequestration potential varied substantially between 0.4 and 2.5 t CO2eq per year and hectare. The modelling approach showed a good fit for apples and wild cherries and – after (re)calibration with local data – also for poplars and sour cherries. Tree mortality was up to 20 % in the first years but if replaced, this did not influence the overall outcome after ten years. Farmers' evaluations differed, depending on the motivation of individual farmers. They changed only slightly with time, indicating that their expectations had been realistic. The study highlights the usefulness of long-term empirical data for model calibration and of monitoring farmers' satisfaction. Realistic model predictions and management of farmers' expectations will facilitate the implementation of agroforestry
Künstliche Intelligenz in der Hochschulbildung : Herausforderungen und Lösungsansätze am Beispiel eines Masterstudienganges in Business Administration
No full textAcademID: 45343Der Einsatz von Künstlicher Intelligenz (KI) verändert die Hochschulbildung grundlegend. Studierende können KI zur Unterstützung beim Schreiben wissenschaftlicher Arbeiten, zur Inhaltserstellung und Präsentationen nutzen. Auch Dozierende profitieren, indem sie KI für die Analyse von Arbeiten, Vorbereitung von Vorlesungen oder das Erstellen von Reports einsetzen. Dabei entstehen jedoch Herausforderungen, insbesondere in Bezug auf ethische Fragen, Datenschutz und faire Bewertung der Eigenleistung. Zudem führt die Einführung von KI in der Bildung zu digitaler Ungleichheit, da nicht alle Studierenden und Dozierenden über die gleichen Kenntnisse und Zugang verfügen. Lösungen können in umfassenden Schulungen für beide Gruppen und einer Anpassung der Leistungsnachweise bzw. ihrer Bewertung bestehen. Die Finanzierung zusätzlicher Kosten für KI-Schulungen und technische Infrastruktur stellt eine wesentliche strukturelle Barriere dar. Das wissenschaftliche Schreiben, in der alle wichtigen Kompetenzen gleichermassen abgefragt werden, bleibt weiterhin wichtigster Bestandteil der Lehre. Deshalb soll der Eigenbeitrag der Studierenden trotz Einsatz von KI in wissenschaftlichen Arbeiten bewertet werden
Status quo und Herausforderungen des Datenschutzes und der Datensicherheit in Schweizer Gemeindeverwaltungen : eine empirische Analyse
No full textDie fortschreitende Digitalisierung bringt sowohl Chancen als auch Risiken für Schweizer Gemeinde- und Stadtverwaltungen mit sich. Besonders der Datenschutz und die Datensicherheit stellen zentrale Herausforderungen dar. Aufgrund der zunehmenden Digitalisierung in öffentlichen Verwaltungen steigt das Volumen an erfassten und verarbeiteten Daten kontinuierlich, was gleichzeitig das Risiko für Datenschutzverletzungen erhöht. Der vorliegende Artikel untersucht den aktuellen Stand der Datenschutz- und Datensicherheitsmassnahmen in den Deutschschweizer Gemeinde- und Stadtverwaltungen, um den Handlungsbedarf zu identifizieren und Empfehlungen abzuleiten. Die zentrale Forschungsfrage dieses Papers lautet: In welchen Bereichen besteht der grösste Handlungsbedarf für Deutschschweizer Gemeinden in Bezug auf die Informationssicherheit? Um diese Frage zu beantworten, wurden drei Unterfragen formuliert: Welche sind die grössten Bedrohungen und Risiken für die Informationssicherheit? Inwiefern beeinflusst die Grösse einer Gemeinde die Fähigkeit, Datensicherheitsmassnahmen umzusetzen? Welche Rolle spielen Schulungen und Bewusstseinsbildung für die Mitarbeiter bei der Verbesserung der Informationssicherheit? Die Untersuchung basiert auf einer quantitativen Online-Umfrage unter den Deutschschweizer Gemeindeverwaltungen. Die erhobenen Daten wurden mittels deskriptiver Statistik analysiert, um einen umfassenden Überblick über die implementierten Datenschutz- und Datensicherheitsmassnahmen zu erhalten. Die Ergebnisse der Untersuchung zeigen, dass insbesondere kleinere Gemeinden im Bereich der Informationssicherheit Unterstützung benötigen. Während technische Schutzmassnahmen weit verbreitet sind, mangelt es teilweise an der Umsetzung strategischer und operativer Massnahmen. Die Analyse verdeutlicht zudem die Notwendigkeit, die bestehenden Massnahmen zu intensivieren und die Sensibilisierung der Mitarbeiter zu verbessern, um die Datensicherheit zu gewährleisten. Darüber hinaus gibt es einen Mangel an umfassenden Schutzbedarfsanalysen und Notfallkonzepten, obwohl 70% der Verwaltungen allgemeine Richtlinien zur Datensicherheit umgesetzt haben.
Advancing digitalization brings both opportunities and risks for Swiss local administrations. Data protection and data security in particular pose key challenges. As digitalization increases, the volume of collected and processed data grows, which also raises the risk of data protection violations. This paper examines the current state of data protection and data security measures in Swiss-German local administrations to identify the need for action and derive recommendations. The central research question is: Which subject areas have the greatest need for action among Swiss-German municipalities and cities regarding information security? To answer this, three sub-questions were formulated: What are the biggest threats and risks to information security? To what extent does the size of a municipality influence its ability to implement data security measures? What role do training and awareness-raising for employees play in improving information security? The study is based on a quantitative online survey of Swiss-German local administrations. The data was analyzed using descriptive statistics to obtain an overview of implemented data protection and security measures. The results show that smaller municipalities in particular need support in information security. While technical protection measures are widespread, strategic and operational measures are sometimes lacking. The analysis also highlights the need to intensify existing measures and improve employee awareness. Furthermore, comprehensive protection requirement analyses and emergency concepts are often missing, although 70% of administrations have implemented general data security guidelines
Learning from safety science : designing incident reporting systems in cybersecurity
No full textDespite all the technical approaches to monitoring threats and detecting incidents, manual incident reporting is critical at all organizational levels of cybersecurity. However, in its current state, reporting suffers from challenges such as underreporting, lack of reporting channels, and uncertainty about what should be reported. The phenomenon of incident reporting itself is not clearly defined and occurs in different facets, from reporting phishing emails to the IT department to reporting vulnerabilities to national authorities. This makes it difficult to design effective socio-technical incident-reporting systems (IRS) according to overarching principles. This review article addresses these challenges by drawing on insights from the field of safety, where IRS are well-established. We find that a broad range of events is reported, various reporting channels on different organizational levels exist, and key design factors of successful IRS have emerged. Based on these lessons from safety, we propose a taxonomy for cybersecurity reporting that includes noncritical events, such as near misses, and latent factors, such as weak security controls. We suggest that also in cybersecurity new reporting channels can be established, e.g. for reporting of noncritical events to nonpunitive supra-organizational bodies or for employee reporting. When designing IRS for cybersecurity, factors such as case-based learning, voluntariness, impunity, independence, and feedback should be taken into account in order to encourage reporting
Möglichkeiten und Grenzen Künstlicher Intelligenz in der umfassenden Finanzplanung
No full textDie Künstliche Intelligenz bietet für die Finanzdienstleistungsindustrie enorme Geschäftspotenziale. Insbesondere bei beratungsintensiven Dienstleistungen, wie der umfassenden Finanzplanung, könnte die generative Künstliche Intelligenz sowohl Einspar- als auch Geschäftspotenziale generieren. Im Rahmen eines Praxisprojekts wurde dieses Potenzial in der umfassenden Finanzplanung eruiert. Der durchgeführte Workshop lieferte die Erkenntnisse, dass nebst technologischen Überlegungen vor allem auch die strategische und betriebswirtschaftliche Verankerung der angebotenen Dienstleistung wichtig ist. Herausforderungen, die bereits heute in der Beratung bestehen (Stichwort: Datenbeschaffung), werden durch den Einsatz von Künstlicher Intelligenz eher noch größer. Die Untersuchung zeigte Chancen bei der Unterstützung der Finanzplanungsprofis auf, kommt aber zum Schluss, dass die Chancen einer autonomen Selbstberatung durch die Endkundinnen und -kunden mithilfe eines Chatbots in keinem positiven Verhältnis zu den Risiken stehen
Carers in focus : ongoing study on technological solutions to support stroke rehabilitation in Switzerland
No full textBackground: Stroke often results in long-term physical, cognitive, and emotional impairments, with care typically provided by informal caregivers, such as family members or friends, who often lack prior experience in this role. These informal caregivers face various challenges, including inadequate preparation and education and limited support, which can lead to physical, emotional, and psychological strain (1). As the number of people living with chronic stroke increases, the need for targeted informal caregiver education has become evident, yet current resources remain limited, require caregivers to attend in person classes at a given time and place. Caregiver-centered approaches and technology-based educational offers, offer a promising way to address these gaps (2).
Aim: This project explores the needs of informal caregivers of people with stroke in Switzerland, aiming to identify key topics for education needs and preferred methods of information delivery. The findings will contribute to developing a tailored Massive Open Online Course (MOOC) for informal caregivers of people living with stroke.
Methods: In this ongoing research project, we plan to collect data through a multilingual online questionnaire. The survey was developed based on qualitative research (1) and literature, targeting informal caregivers in Switzerland. It will assess caregivers’ needs with regards to education about stroke, baseline burden levels (3), and attitudes toward technology. We completed pretesting to ensure usability and data collection is currently underway. We will use descriptive statistics to summarize responses, logistic regression to explore key relationships between caregiver characteristics and support needs and thematic analysis for open-ended responses.
Results & Implications: The study will offer insights into caregiver needs and guide the development of a technology-based educational tool. Results will be presented in May 2025
Off-grid water treatment and reuse of rain- and greywater on household level : pilot operation at the example of KREIS-Haus, Switzerland
No full textThe challenge of urban water scarcity, accentuated by climate change, calls for innovative solutions like KREIS-Haus – an inhabited living lab in Feldbach, Switzerland, demonstrating a decentralized, closed-loop water system. The aim of this study was to evaluate the technical and economic viability of KREIS-Haus's water system, integrating rainwater harvesting and greywater treatment and reuse within a single household setting to achieve water self-sufficiency. Weekly samples were taken pre- and post-treatment of both rain- and greywater treatment and analysed for a set of abiotic and microbial parameters. The effectiveness of the two treatment systems resulted in satisfactory removal rates for COD, BOD5, turbidity and microbial parameters in greywater, however moderate removal rates in rainwater treatment. The treated rain- and greywater was evaluated for its suitability in domestic applications, including drinking, showering, laundry, toilet flushing, and irrigation. The results indicated general adherence, although occasional exceedances highlighted the need for continuous monitoring and system adjustments. Throughout the experimental period, the house maintained complete water self-sufficiency, relying exclusively on rainwater without the need for external water sources. The study also examined the water system's energy usage, resulting in a greater consumption than global averages for wastewater treatment, but with room for improvement through optimization. A cost analysis positions the system as economically competitive, but only when the mandatory connection fees were excluded, thereby underscoring the influence of regulatory frameworks on the adoption of such water systems
