Authors - Dzhansel Abtula, Stanka Hadzhikoleva, Emil Hadzhikolev, Iliana Ivanova, Elitsa Dubarova Abstract - The present study examines the potential of generative language models as tools for discourse analysis of war-related language in media texts. The study is based on a corpus of texts processed by three generative AI models using an identical prompt that defines a multi-stage analytical procedure. This procedure includes the extraction of war-related lexical units, semantic classification, functional analysis of evaluative and ideological features, compilation of a thematic glossary, analysis of metaphors, identification of discursive strategies, and genre determination of the texts. The generated analyses are evaluated through a structured questionnaire based on a five-point Likert scale, completed by an expert with an academic background. The study aims to systematically assess and compare the quality of discourse analyses produced by different language models under controlled conditions. The results indicate that all models generate structurally coherent and terminologically consistent analyses, but differ significantly in interpretative depth and contextual sensitivity. These findings support the need for a hybrid approach that combines automated analysis with human expertise to ensure accurate interpretation of implicit meanings, ideological nuances, and context-dependent discourse features.
Authors - Makomborero Murwira, Dane Brown Abstract - The proliferation of realistic synthetic speech poses significant threats to information integrity and public safety through financial fraud and misinformation campaigns. While traditional countermeasures based on Gaussian Mixture Models have proven effective against earlygeneration deepfakes, they struggle to generalise to sophisticated attacks produced by contemporary neural synthesis. This paper investigates enhanced audio deepfake detection through Light Convolutional Neural Networks (LCNNs) combined with Linear Frequency Cepstral Coefficients (LFCCs) and advanced training strategies. This study systematically evaluates the impact of margin-based loss functions (Cosface and A-Softmax) and FreqAugment data augmentation on model robustness and generalisation capability. Validated on the ASVspoof 2019 Logical Access dataset, the optimised LCNN model achieves an Equal Error Rate (EER) of 5.40% on the evaluation set containing thirteen unseen attack types, representing a 33% relative improvement over the baseline LFCCGMM countermeasure (8.09% EER). The combination of Cosface loss with FreqAugment demonstrates superior performance compared to ASoftmax configurations, reducing false negatives substantially. Per-attack analysis reveals robust performance across diverse spoofing techniques, though vulnerabilities to advanced neural waveform manipulation methods remain. The proposed framework provides a practical, deployable solution for audio deepfake detection in real-world security-critical applications.
Authors - Ida Bagus Dwipayana , Naniek Utami Handayani, Singgih Saptadi Abstract - The global construction sector is under increasing pressure to shift towards sustainable operational models, particularly in heavy materials processing like asphalt production. This study provides empirical, case-studybased insights regarding sustainable procurement practices at the Asphalt Mixing Plant (AMP), run by PT. Jasamarga Tollroad Maintenance in Karawang, West Java, Indonesia. Utilizing six semesters of the Environmental Management Monitoring Effort (UKL-UPL) data from Semester 1 (S1) 2021 to S1 2025, this research assesses the environmental performance of three integrated innovations: (1) a gas-fired burner system substituting for fuel oil, (2) multi-layered noise reduction technologies, and (3) collection and water vapor-based treatment of stone dust to optimize novelty soil fertility systems. Stack emissions analysis found significantly lower levels of CO at 184 mg/m3 and NOx at 212 mg/m3 than Indonesian regulation enables KEP-13/MENLH/3/1995, while ambient air was consistently compliant during monitoring over nine periods. Outdoor sound levels in the upwind and downwind locations were 65.6 dB (A) and 56.8 dB (A), respectively, both below the exposure limit of 70dB(A) specified by SK Men LH KEP-48/MenLH/11/1996, Sensitivity analysis between four fuel-technology scenarios shows a combination approach achieving a 61.9% CO emission index reduction, 55.4% NOX index reduction, and SO2 index is reduced by 78%, relative to diesel fuel basis. This study finds a significant research gap with respect to operationalizable sustainable procurement criteria combining air quality, noise, and waste valorization in an integrated framework at the AMP operational level. The results offer theoretically underpinned and empirically verified procurement criteria, which are directly applicable to the international construction industry, thus advancing green supply chain management, principles of circular economy as well as pathways towards net-zero construction.
Authors - Cossi Blaise Avoussoukpo, Amara Camara, Babou Dione Abstract - Digital scholarly infrastructures increasingly shape academic visibility by supporting the creation, organisation, dissemination, and discovery of research outputs. Existing studies have advanced important dimensions of this domain, including bibliometric evaluation, persistent identiers, metadata standards, and indexing systems. However, they primarily examine these components independently and provide only a limited understanding of how their interactions inuence visibility in contemporary research ecosystems. This paper introduces the Academic Visibility Pipeline Model (AVPM). This conceptual framework conceptualises academic visibility as a sequence of interconnected stages linking research production, metadata structuring, scholarly indexing, identity and aliation aggregation, dissemination, and discovery. By integrating socio-technical coordination, interoperability, and cascading failure, the AVPM explains how visibility emerges and how upstream constraints inuence downstream outcomes. An illustrative application to the Guinea academic visibility ecosystem demonstrates the explanatory value of the framework by identifying structural bottlenecks and tracing their eects across the visibility pipeline. The proposed framework extends conventional publication- and citation-centred perspectives and provides a process-oriented foundation for institutional strategy, research governance, and digital transformation in emerging research ecosystems.
Authors - Duy Nguyen Ngoc, Hiep Nghia Phan Abstract - Large Language Models (LLMs) have recently shown strong capabilities in automatic text summarization. However, applying these models to lowresource languages such as Vietnamese remains challenging due to limited training resources and language-specific characteristics. In this work, we ex-amine whether prompt optimization can improve Vietnamese summarization quality without modifying model parameters. This paper introduces a Vietnamese-aware Prompt Optimization Framework that refines prompt instructions by combining task-specific guidance, role-based prompting, linguistic con-straints, and iterative feedback. The generated summaries are assessed using both automatic evaluation metrics and human judgments to examine the ef-fectiveness of different prompt designs. We evaluate the proposed approach on a benchmark Vietnamese multi-aspect opinion dataset using several commercial and open-source LLMs, including GPT-4, Claude 3, Gemini 1.5, and PhoGPT, and compare their performance with Vietnamese pre-trained summarization models such as ViT5 and BARTpho. Our experiments show that refining prompts consistently improves summary quality across the evaluated models. In particular, GPT-4 with the optimized prompt achieves an 8.7% increase in ROUGE-L and receives higher human evaluation scores for fluency and factual consistency than the standard prompting setting.
Authors - Tien-Dao Luu, Viet Truong Xuan, Doan Hoang Phuc Nguyen, Nghi Huynh Quang, Do Chau Giang Nguyen, Nghia Nguyen Khoi, Huu Hoa Nguyen Abstract - The Mekong Delta agroecosystem faces compounding climate stressors and acute data fragmentation. While agricultural data exists online, it remains largely unstructured and unverified. This study introduces WikiCrop-AI, an integrated data-processing and machinelearning framework designed to consolidate heterogeneous agronomic information. The architecture comprises three interconnected modules: an Agricultural Notebook utilizing a Retrieval-Augmented Generation pipeline for multi-format data ingestion, a browser-based computational environment (WikiLab) for reproducible workflows, and a client-side analytics module for multivariate clustering. Evaluation of the ingestion pipeline yielded a String Similarity Score of 0.99 for structured text extraction. Furthermore, generative outputs assessed via the Automatic LLMs Citation Evaluation framework achieved average scores of 0.94 for both Citation Recall and Precision, alongside a Claim Recall of 0.92, indicating reliable knowledge synthesis under the tested conditions. A hierarchical clustering case study on 22 soybean cultivars further illustrates the platform’s utility in supporting non-programming local agronomists. Ultimately, WikiCrop-AI provides a decentralized infrastructure to translate scattered digital resources into actionable, verified agricultural intelligence. The complete open-source code for the WikiCrop-AI ecosystem is publicly accessible on GitHub at: https://github.com/mekonglab-vn/ WikicropAI.
Authors - Harshala Shingne, Shwetambari Borade, Dhanashree Hadsul, Aditya D. Nandgirwar, Pranali Pawar, Rupali Vairagade Abstract - Increasing proliferation of networked systems have compounded the necessity to seek effective and non-invasive intrusion detection solutions. The conventional intrusion detecting systems (IDS) are mainly centralized into data aggregation scheme that introduces essential constraints pertaining to data exposure, scalability, and robustness of the system itself. Partially in reaction to this, this paper presents a privacy conscious federated intrusion detection design that allows collinear model training by many network participants in the absence of exchanging raw data. This framework exploits federated learning to create a global intrusion detecting model by continually aggregating local-trained updates, thus retaining the data locality and ownership. In order to achieve high privacy assurances, there are inbuilt secure aggregation mechanisms and perturbation-based mechanisms that accomplish this by avoiding the leakage of sensitive information during model sharing. Additionally, an adaptive-aggregating strategy is proposed that can effectively manipulate non identically distributed data of the participants, as well as improving the generalization process of the global model. Extensive testing on test sets of benchmark intrusion detection has shown that the proposed framework has very high detection rates and much less privacy risk and communication overhead than the traditional centralized techniques. The findings confirm that the framework has the capability of offering a scalable, secure and efficient intrusion detection solution in distributed networks.
Authors - Shashank Mallesh, Anithadevi M D, Srinidhi G A, Chandana Sreenivas Abstract - Hospital surge capacity management remains a critical challenge in healthcare systems, particularly during pandemic events. This research presents a novel Adaptive Multi-Objective Capacity Management (AMCM) framework that integrates Long Short-Term Memory (LSTM) networks with Multi-Objective Particle Swarm Optimization (MOPSO) to optimize bed allocation, staffing schedules, and equipment distribution. The framework simultaneously minimizes patient wait times, operational costs, and resource wastage while maximizing bed utilization efficiency. Comprehensive evaluation against state-of-the-art algorithms including Support Vector Regression (SVR), Random Forest (RF), Gradient Boosting (GB), and traditional Mixed Integer Linear Programming (MILP) demonstrates superior performance across all metrics. The proposed method achieves 94.7% bed utilization accuracy, reduces emergency department wait times by 42.3%, and decreases surge-related costs by 38.9% compared to conventional approaches. Validated on real-world COVID-19 hospital data spanning 18 months across five major health systems, the AMCM framework provides healthcare administrators with an intelligent decision support system for proactive capacity planning and dynamic resource allocation.
Authors - Nwagu Chima Ajanwachuku, Onyemaobi Bethram Chibuzo, Nwafor Franca Amaka, Divine Nnodim Oluchi Abstract - Tertiary institutions across the world are now adopting artificial intelligence-based academic detection systems to aid in detecting different forms of academic misconduct. For these detection systems, there is more focus on technical performance metrics such as detection accuracy, precision and recall and little focus on whether these systems validly measure the complex construct of academic misconduct. This study aims to examine how AI-based academic integrity detection systems operationalise, measure, and validate academic misconduct in higher education, focusing on construct operationalisation, measurement accuracy, and construct validity, through a systematic literature review. We conducted a systematic review and retrieved articles from ACM Digital Library, IEEE Xplore, Web of Science, and Google Scholar. Of 793 articles, 56 were selected using the PRISMA framework, and the findings were synthesised narratively. The 56 studies focused on plagiarism detection, AI-generated text detection, authorship verification, behavioural monitoring, biometric authentication, and multimodal detection systems. Across all the studies we considered, detection systems mainly measured observable digital signals. 55 of 56 studies showed evidence of construct misalignment between the measured signal and the claimed misconduct construct. Most studies we considered treated similarity as plagiarism, AI-generated probability as dishonesty, and behavioural anomalies as cheating, even though these signals could not capture intent, differentiate between acceptable collaboration and collusion, or account for disclosure practices or alignment with institutional policy. Also, we observed that most studies validated detection systems using technical metrics such as accuracy, precision, recall, and F1 score, and just a few directly addressed construct validity, bias, or robustness.
Authors - Dang Trung Thanh, Nguyen Huynh Anh Tuyet Abstract - The objective of this project is to apply GPS devices in combination with GIS software to build image interpretation keys for spatial data management. The research content includes: collecting satellite imagery data and documents; conducting field surveys and collecting GPS coordinates for 149 sample points; and building image interpretation keys for geographical objects. The main research method combines GPS field surveys, remote sensing image interpretation, and the application of GIS software such as QGIS and Google My Maps to process, analyze, and manage spatial data. In addition, the project utilizes methods of document collection, statistics, and comparison to ensure the accuracy and scientific validity of the research results. The project results: A set of image interpretation keys was developed for several key geographical objects in the study area, including: water bodies (27 samples), transportation (33 samples), agricultural land (27 samples), residential construction land (28 samples), and vacant land (34 samples). The research has contributed to demonstrating the effective application of GPS combined with GIS and remote sensing in surveying, mapping, and managing geographic information on current land use. The research and development direction is: Integrating artificial intelligence (AI) and machine learning (Deep Learning) into the image interpretation process
Authors - Dang Trung Thanh, Nguyen Huynh Anh Tuyet Abstract - This study evaluates the implementation of the 2024 Land Law regarding land-use conversion in Thuan An Ward, Ho Chi Minh City, Vietnam. The research aims to assess the changes introduced by the new legal framework and examine its practical impacts on land-use conversion procedures at the local level. The study employed a combination of document analysis and a questionnaire survey of 100 respondents, including local government officials and citizens. The collected data were analyzed using descriptive statistics and comparative methods. The results indicate that the 2024 Land Law has improved the landuse conversion process by simplifying administrative procedures, reducing processing time, and increasing transparency. Survey findings show that most respondents considered the new regulations easy to understand, the processing time efficient, and the procedural costs reasonable. Overall public satisfaction increased from 65% under the 2013 Land Law to 85% under the 2024 Land Law. Nevertheless, several challenges remain, including incomplete digital land databases, limited public understanding of some legal provisions, and issues related to the implementation of market-oriented land pricing. These findings provide practical evidence for improving land-use conversion management and support the effective implementation of the 2024 Land Law at the local level.
Authors - Rebecca Hufkie, Dane Brown Abstract - Protein structure determines function, yet experimental determination methods remain costly and slow. This study presents an optimised transformer-based system for predicting protein sidechain angles and reconstructing 3D structures directly from sequence data. Trained on the SidechainNet CASP12 dataset comprising 25,044 proteins, the systematically refined model achieves 0.244 radians RMSE for angle prediction and 1.413 ĚŠA RMSD for structural accuracy, representing a 70% improvement over the baseline. Incorporating backbone angles, secondary structure, and evolutionary information reduces RMSD from 1.861 ĚŠA to 1.413 ĚŠA compared to sequence-only inputs. On challenging CASP12 free-modelling targets, the system scores 84 to 86 GDC. This performance is competitive with leading methods while maintaining computational efficiency through single-sequence prediction without multiple sequence alignment generation. Results indicate that specific architectural choices, including deeper networks, GELU activation, and increased embedding dimensions, combined with robust dropout and weight decay, enable highly accurate structure prediction from limited training data. This demonstrates that carefully constrained models can capture complex biological folding patterns efficiently without massive computational overhead.
Authors - Daniel H. M. Marques, Luiz A. P. Silva, Andson M. Balieiro, Mohammed B. Alshawki, Alexandre J. R. Serres, Dalton C. G. Valadares Abstract - The evolution towards 5G and Beyond (5G/B5G) standardization, and even the development of 5G, can be accelerated by using accessible, high-fidelity emulation environments to validate emerging network architectures. For instance, Network Slicing is an aspect that can especially benefit from these tools. However, defining an emulation platform that is compatible with research objectives can be challenging. So, this work aims to compare two different Mobile Network emulation setups: one using Open5GS to emulate the Core Network and UERANSIM for the implementation of the Radio Access Network (RAN) and User Equipment (UE), and another using an OpenAirInterface (OAI) End-to- End implementation. Furthermore, this work aims to fill relevant gaps in the academic literature, addressing implementation obstacles at a granular level and the architectural trade-offs necessary to stabilize these environments. To that end, we identified and resolved operational friction points such as kernel-level GPRS Tunneling Protocol User Plane (GTP-U) conflicts, slice identity alignment, and Physical layer (PHY) timing sensitivities in virtualized radio frequency simulators through two deployment frameworks. Our analysis shows that, while the Open5GS/UERANSIM stack offers better agility for Core Network prototyping, OAI offers a more flexible and feature-rich framework, suitable for researching advanced RAN features, albeit with greater configuration complexity. By documenting troubleshooting protocols and architectural comparisons, this work serves as a practical guide for researchers migrating from 5G simulation to 5G/B5G-ready emulation test environments.
Authors - Aditya D. Nandgirwar, Deepika Burte, Rashmi Malvankar, Rupali Vairagade, Shwetambari Borade, Sandeep M. Chitalkar Abstract - Another significant development in intelligent system development is the notion of Agentic Artificial Intelligence (Agentic AI): as passively generative models continue becoming acts-oriented entities capable of perceiving their environment, thinking about goals, and planning and executing more complex tasks without necessarily involving humans. Unlike the traditional artificial intelligence systems, whose main input is the fixed input, agentic AI systems respond to objectives, are adaptive in their decision-making and are able to interact with other tools, environments and fellow agents. Thus agentic systems are increasingly being used in diverse technical disciplines, including software development and cybersecurity, healthcare, finances, robotics and enterprise automation. The paper provides a thorough description of agentic AI, its theoretical basis, architectural design components, implementation plans, security issues and applications. We take a look at intelligent agent development and explain fundamental concepts in the designing of intelligent agents such as the reasoning, the strategy to plan, the management of memory, and collaboration of intelligent agents. Moreover, we examine leading agentic AI systems and platforms that facilitates development and coordination of autonomous systems.
Authors - A.N. Gachahi, L.W. Gachahi Abstract - Antennas are fundamentally important in all modern wireless communication systems. Their design directly influences the performance, efficiency, and size of wireless devices. A core limitation of existing antenna designs is their reliance on spatial resonance where antenna size is proportional to a fraction of the wavelength of the signal they are intended to transmit or receive. Here, we present the feasibility of using single-layer capacitors (SLCs) as radiating elements for antenna applications. The SLC approach exploits displacement currents and the temporal resonance inherent to capacitive structures. A theoretical framework is established using Maxwell’s equations and circuit-level analysis to explain how electromagnetic fields arise around SLCs. An experimental setup using an array of 90 ceramic capacitors is constructed, and a magnetic field sensor is used to measure radiated fields across a range of frequencies. The antenna is then modeled in MATLAB, and simulations are performed to evaluate radiation patterns and impedance characteristics. Results confirm field generation consistent with theoretical predictions. Integration into ESP32-C3 Wi-Fi modules is demonstrated. Impedance mismatch is identified as a key limitation and addressed through a resistive matching network, improving signal consistency under varying distance conditions. The theoretical, experimental and simulation results in this set-up confirm that SLCs can serve as efficient, compact antenna elements, with practical implications for RF systems where size, cost, and integration constraints are critical.
Authors - PARVIZ FIRUDIN OQLU KAZIMI, KAZIM ASAD OQLU KAZIMLI Abstract - Objective: The targeted modeling of social information in global, regional, and local projects, and the use of information influence for both progressive and aggressive purposes, constitutes an activity in the field of information engineering. It is important to study the multifaceted and complex scientific and theoretical foundations of this activity and discuss them in a broad academic community. Theoretical Foundations: It is inappropriate to limit information engineering to technical, technological, and software issues; it is important to study the scientific, theoretical, and experimental aspects of information influence at various levels using modern technologies. Research Methods: This study presents considerations regarding methods for targeted information modeling in education, culture, and information-intensive fields. The proposed considerations are intended to standardize a number of processes, identify aggressive elements in some areas, and, in some cases, consider innovative information modeling. The proposed concept can be compared with a number of sociological analytical models. However, for the first time, it is proposed that the essence of information, the study of the aspects of thesauri influence, and the application of modern technologies will gain greater relevance. This study addresses the application of our theoretical work, known as "information
Authors - Vishakha Shinde, Himangi Pande Abstract - The rapid growth of digital communication platforms, cloud multimedia sharing and AI-driven visual systems has increased the demand for secure image ownership verification and multimedia authentication. Conventional watermarking approaches often suffer from limited robustness against compression, geometric distortion and adversarial attacks. Recent advances in deep learning and cryptographic protection mechanisms have enabled the development of intelligent hybrid watermarking frameworks with improved robustness, adaptive embedding and secure ownership verification. This paper presents a comprehensive review of secure image watermarking techniques integrating deep learning architectures and cryptographic security schemes. The study analyzes CNN-, autoencoder-, GAN- and diffusion-based watermarking frameworks along with encryption-assisted watermark embedding strategies, attack resilience mechanisms and lightweight watermarking systems for edge– IoT environments. Benchmark datasets, performance metrics, loss functions and quantitative comparisons of existing frameworks are also discussed. The comparative analysis indicates that hybrid deep learning–cryptographic watermarking methods significantly improve robustness, authentication reliability and resistance against signal-processing, geometric and adversarial attacks. However, computational complexity, scalability and real-time deployment constraints remain major challenges. Finally, the paper discusses future research
Authors - Meghali Kalyankar, Prashant Lahane Abstract - As deepfake generation technologies have rapidly advanced, establishing authenticity for multimedia content on digital platforms has become a major challenge[1]. Current deepfake detection approaches primarily rely on unimodal analysis and face challenges in encoding joint AV inconsistencies, temporal consistency, and adversarial attacks. To overcome these problems, a novel Multimodal Attention and Adversarial Deepfake Network (MMAD-Net) framework for robust audiovisual deepfake detection is proposed. The proposed solution adopts a framework that combines fine-grained visual feature extraction provided by VideoMAE v2 [2], audio representation learning provided by HuBERT [3] and temporal dependency modeling provided by TimeSformer [4] to model the fine-grained spatial and temporal inconsistency in manipulated media. In addition, a Multimodal CoAttention Transformer (MCAT) is used for better cross-modal interaction between audio and visual streams, and a hybrid HOA-COA optimization scheme optimizes discriminative feature representations to ensure better feature separation and remove redundancy. Adversarial Consistency Training (ACT) is embedded in the learning process to enhance adversarial robustness against adversarial perturbation and unseen adversarial manipulation. FakeAVCeleb and Celeb-DF are used for testing the proposed model with several performance metrics. Experimental results show that MMAD-Net can provide stable and general detection performance while maintaining a high level of robustness for current multimodal deepfake detection methods.
