16168 research outputs found
Sort by
Dirbtiniu intelektu pagrįstos strategijos, skirtos spręsti natūralios kalbos apdorojimo iššūkius mažai išteklių turinčioms kalboms.
No full textThis dissertation explores AI-driven solutions to advance Natural Language Processing (NLP) for low-resource languages, with a primary focus on Amharic. While high-resource languages benefit from vast linguistic resources and tools, languages like Amharic lack annotated datasets and computational frameworks, limiting the development of core NLP applications. This research addresses those challenges by developing classification models, implementing transformer-based embeddings, and applying innovative data augmentation methods. It also integrates Explainable AI (XAI) techniques to enhance transparency and trust in model predictions. Key applications examined include sentiment analysis, intent recognition, cyberbullying detection, deepfake recognition, and part-of-speech tagging. The study demonstrates that tailored NLP methods not only improve performance for Amharic but can also be generalized to other underrepresented languages. Overall, this work contributes to the inclusivity and robustness of AI technologies in linguistically diverse digital spaces
Optimizing power sources for smart building sensors: a comparative study of LiSOCL2 batteries under controlled discharge profiles /
No full textLithium thionyl chloride (LiSOCl2) batteries are pivotal in enabling long-term, maintenance-free operation of smart building sensor networks due to their superior energy density, exceptional shelf life, and reliability in extreme environments. These attributes make them particularly suitable for powering a diverse array of embedded electronic devices within smart infrastructure—including wireless HVAC sensors, high-voltage direct current sensing units, low-power IoT nodes, security and occupancy detectors, structural health monitors, and adaptive lighting or ventilation controllers. This study presents an empirical evaluation of four leading LiSOCl2 battery brands—EVE, Saft, TEKCELL, and TADIRAN—to assess their real-world performance under varying discharge currents, with direct implications for power circuit design in smart building sensor networks. Despite similar datasheet specifications, our findings reveal substantial discrepancies in actual performance, capacity retention, and degradation characteristics. This highlights the need for empirical validation in battery selection, beyond nominal manufacturer ratings, especially when deployed in systems requiring sustained ultra-low power draw over multi-decade lifespans. These discrepancies found underscore the need for accurate battery characterization in energy budgeting, adaptive duty-cycling, and intelligent load management strategies. The findings also inform battery selection for hybrid systems incorporating energy harvesting, enabling sustainable, maintenance-free sensor deployment in energy-optimized building environments. By aligning real-world battery behavior with architectural design choices in sensor systems, our results support the deployment of scalable, sustainable sensor networks in smart buildings. These networks can operate autonomously for 10–40 years, reducing lifecycle maintenance costs and material waste, thereby advancing goals in energy-efficient, digitally optimized building design
Effect of combined heat and mechanical processing on the hardness and wear resistance of X160CrMoV12 tool steel /
No full textThis study investigated the effect of cold plastic deformation at Bridgman anvil chamber temperature on the hardness and wear resistance of X160CrMoV12 steel using hardness testing, X-ray diffraction (XRD), abrasive grinding wear (AEMW) testing, optical examination, and scanning electron microscopy (SEM). Three batches of samples were prepared for the experiment: I – hardened, II – hardened and then tempered at 600°C for 1.5 hours, III – hardened and then plastically deformed. The samples were hardened at three temperatures: 1100, 1150, and 1200 °C. The highest amount of retained austenite, reaching 69.02%, was observed when hardening at 1200°C, while at lower temperatures, 17.36% and 38.14% were formed, respectively. After hardening (batch II), the amount of retained austenite decreased proportionally by approximately 7 times for each hardening temperature. The effect of plastic deformation (batch III) is observed by analysing the hardness of samples from the surface to the depth, reaching an average hardening depth of 0.08 mm. To check how well it holds up to wear, the surfaces of three test batches were tested using an abrasive grinding test with a load of 5N. Hardened and plastically deformed specimens showed greater resistance to abrasion than hardened and tempered specimens. The results confirmed that the optimal hardening temperature for achieving maximum wear resistance of this steel is 1100°C
Machine learning-based measurement forecasting approach for smart agriculture /
No full textThis work explores whether a low-resolution thermal camera can estimate three discrete sensor measurements on a resource-constrained IoT node. Correlation analysis showed that individual thermal pixels correlate strongly with air temperature, negatively with relative humidity and positively with light intensity. Three lightweight regressors VGG CNN, ViT-Tiny and CvT-Tiny were trained from 1 053 single channel 120 x 160-pixel thermal frames to estimate sensor measurements. Experimental tests confirmed the CNN superiority as it achieved RMSE of 2.29 °C and R² of 0.978 (estimating air temperature), RMSE of 0.075 %RH and R² 0.897 (estimating relative air humidity) and RMSE of 0.059 lux and R² of 0.924 (estimating light intensity), outperforming ViT-Tiny and CvT-Tiny on humidity and light intensity estimation. The findings demonstrate that convolutional models remain critical for lightweight and accurate environmental measurement estimation in edge deployments
Introduction to the minitrack on Human-Robot Interaction and Collaboration.
No full textThis minitrack explores the forefront of Human-Robot Interaction (HRI), with a focus on the ongoing evolution toward seamless Human-Robot Collaboration (HRC). By addressing the latest breakthroughs in methodologies, technologies, and research, this minitrack aims to shed light on the dynamic and rapidly evolving relationship between humans and robots. Experts will present state-of-the-art advancements that are driving this transformation, fostering deeper integration and cooperation between humans and robotic systems
Widening the horizon of anthropocentric interior design towards meaningful human-plant interaction /
No full textHuman-plant interaction is an issue with various dimensions. Humans are inextricably linked to plants and depend on them for their survival. However, they have also developed an anthropocentric and animal-centric attitude and tend to ignore the contribution of plants to their well-being and even to overlook the presence of plants. In recent decades, with the growing body of research on human dependence on nature and positive influence of interaction with nature on human physical and psychological health, a countertrend has emerged: design trends aiming to increase the exposure of urban dwellers to nature and its elements in every step of their daily lives. This study hypothesizes that interiors with integrated indoor plants, designed with appropriate ethical, aesthetic and psychological considerations, can create preconditions and opportunities for more meaningful human-plant interactions and broaden the ethical horizons of interior design towards non-anthropocentric cultural attitudes
Systematic variation of the acceptor electrophilicity in donor-acceptor-donor emitters exhibiting efficient room temperature phosphorescence suited for digital luminescence /
No full textPurely organic materials showing efficient and persistent emission via room temperature phosphorescence (RTP) allow the design of minimalistic yet powerful technological solutions for sensing, bioimaging, information storage, and safety applications using the photonic design principle of digital luminescence. Although several promising materials exist, a deep understanding of the underlying structure-property relationship and, thus, development of rational design strategies are widely missing. Some of the best purely organic emitters follow the donor-acceptor-donor design motif. In this study, the influence of the acceptor unit on the photophysical properties is systematically analyzed by synthesizing and characterizing variations of the RTP emitter 4,4'-dithianthrene-1-yl-benzophenone (BP-2TA). The most promising candidates are also tested in programmable luminescent tags as a potential application field for information storage. While no significant influence by the electrophilicity index of the acceptor moiety on the RTP emission is observed, the results support the design of molecules with pronounced hybridization as obtained for the newly synthesized emitter demonstrating superior RTP efficiency combined with improved stability
Comparison of methods and results for determining optical characteristics of bismuth ferrite thin films grown by reactive magnetron co-sputtering /
No full textBismuth ferrite (BFO) thin films were deposited by reactive magnetron co-sputtering in a pure oxygen environment at varying substrate temperatures in the range of 475 °C–550 °C in increments of 25 °C. As-deposited films were characterized for their structural, morphological and optical properties using X-ray diffraction (XRD), scanning electron microscopy (SEM), and optical absorption measurements. SEM and XRD analysis showed increase grain growth and crystallinity with increase of substrate temperature. Highest phase purity was observed in the sample grown at 500 °C and higher temperatures showed degraded phase purity with highest content of Bi 2 Fe 4 O 9 and β -Bi 2 O 3 . The optical band gap of BFO films varied in the range of 2.32–2.36 eV. The refractive index n( λ ) was evaluated using Kramers-Kronig relations and the Swanepoel envelope method. Kramers-Kronig relations calculation showed a monotonic increase in n( λ ) as opposed to Swanepoel envelope methods non-monotonic results. Transfer matric models confirmed that the values of n( λ ) obtained by the Swanepoel envelope method align more closely to experimental transmittance spectra despite the samples weakly modulated spectra
What determines energy tax rates in European Union countries? /
No full textType of the article: Research Article AbstractThe modern tax system must support the transition to a carbon-neutral economy. Increasing the environmental tax burden is the most effective measure to achieve this goal. In this context, it is essential to assess the determinants of energy tax rates in European Union countries and understand the crucial conditions for making informed policy decisions. This study contributes to the existing literature by examining the determinants of energy tax rates. It incorporates not only macroeconomic, energy efficiency, and environmental factors, but also indicators of companies’ financial performance. The study analyzes a sample of European Union countries from 2010 to 2020, using fixed effects panel regression analysis. The results indicate a negative relationship between energy tax rates and energy intensity (β = –0.347), the return on equity of non-financial companies (β = –0.058), and investments (β = –0.202). The results indicate that energy tax policies in European Union countries are primarily influenced by incentives related to economic growth, specifically energy consumption (β = 0.389), renewable energy (β = 0.076), trade openness (β = 0.544), and the level of public debt (β = 0.234). The results show that environmental motives are not yet a significant factor in the decision-making to increase energy tax rates. The findings indicate that when determining energy tax rates, national governments must carefully consider the balance between environmental motives and the potential consequences for the financial performance of non-financial corporations and their investments, especially in countries with energy-intensive industries
Decarbonization through supply chain innovation: role of supply chain collaboration and mapping /
No full textThe urgency of reducing carbon emissions has intensified amid escalating climate change concerns. Supply chain innovation practices are increasingly recognized as critical enablers of decarbonization by fostering efficiency, sustainability, and carbon reduction strategies. Against this backdrop, this study examines the role of SCIP in supply chain decarbonization. We also explore how supply chain collaboration and supply chain mapping can play a role in mediating the impact of SCIP, if any, on decarbonization. The study is contextualized in Electrical and Electronics sector of Malaysia. Data were collected through close-ended questionnaire from 156 firms. We employed Partial Least Squares Structural Equation Modeling to analyze these relationships. The results confirm a significant direct impact of SCIP on SCD, underscoring the pivotal role of innovation in sustainability efforts. However, contrary to conventional wisdom, SCC does not significantly mediate this relationship, suggesting that collaboration alone may not directly enhance decarbonization outcomes. In contrast, SC mapping plays a crucial mediating role, highlighting its importance in translating SCIP into effective carbon reduction strategies. These findings provide theoretical contributions to supply chain sustainability literature by distinguishing between collaboration and mapping as enablers of decarbonization. Practically, the study underscores the need for firms to invest in digital supply chain mapping tools to enhance visibility and strategic decision-making for decarbonization. Future research should explore industry-specific variations and the role of emerging digital technologies in strengthening supply chain sustainability