Search Results (2,227)

Sentiment analysis of online game reviews has increasingly become important in understanding player experiences and supporting data-driven game development. However, research in this domain has continuously faced two unresolved challenges: (1) the extreme imbalance between positive and negative feedback, and (2) the inefficiency of existing feature-weighting schemes in capturing sentiment signals embedded in informal gaming discourses. Prior works demonstrated that negative feedback—though a few in number are highly influential—usually contain richer emotional content and longer textual structures; yet, prevailing classification models often perform poorly for these minorities (i.e., negative feedback). Numerous studies explored multimodal imbalance issues, class imbalance in cross-lingual ABSA (Aspect-Based Sentiment Analysis), reinforcement-learning-based architectures for imbalanced extraction tasks, and oversampling strategies like SMOTE (Synthetic Minority Over-sampling Technique) variants. Few investigations specifically addressed imbalanced sentiment classification in the contexts of online game reviews, where user-generated content exhibits unique lexical, structural, and emotional characteristics. To address these gaps, this study integrated TF-IDF (Term Frequency-Inverse Document Frequency), VADER (Valence Aware Dictionary and Sentiment Reasoner) lexicon features, and IGM (Inverse Gravity Moment) weightings with advanced oversampling methods such as ADASYN (Adaptive Synthetic Sampling Approach for Imbalanced Learning) and Borderline-SMOTE to improve the detection of minority sentiment classes. Ensemble models, including XGBoost (Extreme Gradient Boosting) and LightGBM (Light Gradient-Boosting Machine), were further employed to enhance the robustness of imbalance. Using a large-scale dataset of Steam game reviews, the proposed framework demonstrated substantial improvement in identifying negative sentiments, addressing a critical limitation in the existing computational game-analysis literature, and advancing the modeling for detecting the emotion-rich but imbalance-prone user feedback. Full article

Large language models (LLMs) have been widely applied to knowledge graph question answering (KGQA) systems. Recent Text-to-SPARQL studies have demonstrated that generation performance can achieve an F1 score exceeding 90%. Further error analysis has categorized common errors into entity translation errors, entity position errors, and resource description framework (RDF) triple-count errors, with the latter accounting for 24% of all errors. Notably, nearly 90% of RDF triple-count errors occur when the triples involve class entities. Previous research has shown that incorporating prompts can effectively enhance model performance. Based on the results, we predicted whether a question contains a class entity and the number of RDF triples in the corresponding query to reduce RDF triple-count errors in large language models by providing precise task-related information through prompt design. Since both strategies are classification-oriented, two implementation paradigms were established: traditional classification architectures and generative modeling. They were compared in terms of performance. For classification-based architectures, we employed Bidirectional Encoder Representations from Transformers (BERT) and the Robustly Optimized BERT Approach (RoBERTa) to obtain question embeddings for classification. For the generative approach, we adopted the Instruction-Tuned Text-to-Text Transfer Transformer (Flan-T5). Experimental results show that the generative model slightly outperforms conventional classification architectures, indicating that generative approaches can achieve higher prediction accuracy and provide more reliable information without the need for additional complex encoder designs, thereby improving the overall quality of Text-to-SPARQL generation. Full article

Armillaria species hold significant ecological and economic importance and they play a vital role in the growth of traditional Chinese medicine Gastrodia elata (G. elata). In this study, we assembled and compared the mitochondrial genomes (mitogenomes) of six Armillaria mellea (Vahl) P. Kumm. (A. mellea) strains isolated from the main G. elata-producing region of Hanzhong, China. The internal transcribed spacer (ITS) sequencing confirmed that all six strains form a monophyletic clade. Their mitogenomes (120,775 to 120,839 bp) exhibit a highly conserved architecture, each containing 16 protein-coding genes (PCGs), 23 open reading frames (ORFs), 27 tRNAs, and two rRNAs. Codon usage and amino acid frequency were strikingly similar among the six strains, with a strong AT bias. In contrast, comparisons with other Armillaria species revealed marked differences in gene order, repeat structures, and selection pressures. Phylogenetic analyses based on PCGs further resolved the close relationship among the six strains while highlighting distinct molecular variation across species. On the whole, these findings demonstrate that A. mellea strains co-evolving with G. elata maintain a highly uniform mitochondrial genome architecture, suggesting strong purifying selection or recent divergence within this symbiotic population. The pronounced differences from other Armillaria species at the levels of gene arrangement and selection pressure imply that mitochondrial gene rearrangement may have accompanied species diversification in the genus. By providing the first complete mitogenomes of A. mellea from a major G. elata cultivation area, this study not only expands the genomic resources for Armillaria but also establishes a foundation for understanding how mitochondrial variation might influence fungal growth, adaptation, and symbiotic efficiency with G. elata. Full article

Extreme class imbalance is a persistent obstacle for machine learning-driven intrusion detection, as rare but high-impact cyberattacks occur far less frequently than benign traffic in training data. In many real-world cybersecurity datasets, this imbalance becomes extreme, with certain attack types containing a handful of samples, effectively placing the problem in a few-shot learning regime. This paper presents a controlled benchmarking study of Generative Adversarial Network (GAN) objectives for synthesizing minority-class cyberattack data. Using the UWF-ZeekData22 network traffic dataset, each MITRE ATT&CK tactic is framed as a separate binary detection task, and tactic-specific GANs are trained solely on minority samples to generate synthetic attack records. Four widely used GAN variants—Vanilla GAN, Conditional GAN (cGAN), Wasserstein GAN (WGAN), and Wasserstein GAN with Gradient Penalty (WGAN-GP)—are compared under unified training steps and fixed augmentation conditions. The utility of generated data is assessed by evaluating downstream detection performance using five traditional classifiers: Logistic Regression, Support Vector Machine, k-Nearest Neighbors, Decision Tree, and Random Forest. The results indicate that GAN augmentation generally strengthens minority-class detection across tactics and models, reducing false negatives and improving recall consistency, while not systematically harming majority-class performance. However, the effectiveness of each GAN objective varies significantly with data sparsity. Specifically, simpler adversarial objectives often outperform more complex architectures by preserving discriminative feature structure, while heavily regularized models may overly smooth minority-class distributions and reduce separability. Wasserstein-based objectives provide improved training stability, but additional regularization does not consistently translate to better detection performance. Overall, the results demonstrate that in extreme-imbalance settings, GAN effectiveness is governed more by data sparsity and structure preservation than by architectural complexity. These findings establish class-specific generative augmentation as a practical strategy for intrusion detection and provide empirical guidance for selecting appropriate GAN objectives for tabular cybersecurity data under highly imbalanced conditions. Full article

Biomass-derived sulfur-containing hard carbons are promising anode candidates for sodium-ion batteries, but cross-study optimization remains difficult because reported electrochemical performance reflects both synthesis history and incomplete or non-uniform structural characterization. Here, we assembled a focused literature-derived dataset of 101 records from 16 journal articles and compared the predictive value of three information sources: precursor descriptors, process variables, and measured structural descriptors. We further introduced domain-knowledge-guided precursor descriptors to encode interpretable aspects of precursor chemistry and architecture, including lignin-related richness, polysaccharide contribution, volatile tendency, precursor-component coupling, and post-treatment category. In controlled feature-set comparisons, the model combining precursor and process descriptors achieved an R² of 0.59, outperforming the conventional combination of process and structural descriptors (R² = 0.57) and remaining close to the full-information setting (R² ≈ 0.61). Model interpretation further showed that, when structural descriptors were removed, predictive reliance shifted toward precursor and process variables, indicating that accessible upstream descriptors retain a meaningful fraction of the formation-pathway information relevant to sodium storage. These results should be interpreted within this curated sulfur-containing literature space rather than as a universal predictor, but they demonstrate that domain-knowledge-guided precursor encoding can support low-characterization, screening-oriented prediction and experimental prioritization. Full article

This study focused on elucidating the effects of chitosan (Ch), hyaluronic acid (HA), and titanium dioxide nanoparticles (nano-TiO₂) on the physicochemical characteristics of a model bacterial membrane (layer) composed of the phospholipid DPPG (1,2-dipalmitoyl-sn-glycero-3-phospho-rac-(1-glycerol) sodium salt). The membrane was prepared on mica using the Langmuir–Blodgett (LB) technique from an aqueous subphase containing Ch, HA and/or TiO₂. Its surface properties were subsequently characterized by optical profilometry and surface free energy estimation. The nanoscale topography of the DPPG layer provided a biomimetic platform that reflects the organization of bacterial membranes, enabling a precise evaluation of how external agents, such as Ch, HA, and nano-TiO₂, modify the surface’s structural and energetic properties. The results showed that the LB films exhibit mildly heterogeneous topography, which can be attributed to lipid domains with distinct molecular packing densities. Depending on the type of biopolymer employed with TiO₂, distinct topographic architectures of the DPPG monolayers were obtained. Furthermore, the presence of nano-TiO₂ was clearly manifested as a topographic irregularity, while the analysis of hydrophilic–hydrophobic properties revealed a structurally perturbed lipid film. The results provide detailed insight into how these specific molecules (Ch, HA, nano-TiO₂) interact at the molecular level with model bacterial membranes, offering a comprehensive picture of cell–microenvironment interactions. Full article

Meticulous segmentation of medical images requires obtaining both local and global spatial detailed information. The conventional U-Net model excels at local spatial feature extraction through residual convolutional blocks but struggles to capture global features. To resolve this issue, we propose the vision transformer U-NeT (ViTUNet) model framework, which combines the self-attention mechanism of the vision transformer (ViT) to capture global information while maintaining the extraction of local features via U-NeT. This proposed architecture introduces vision transformers to the existing residual convolution blocks in the U-Net encoder path, thereby capturing both local and global features. The decoder path then rebuilds this information into high-quality segmentation maps with accurately highlighted boundaries/edges. This model is utilized to segment carious lesions in bitewing dental radiographs. These images are pre-processed using augmentation, morphological operations, and segmentation to identify the boundaries/edges of the regions of interest (caries/cavity). The proposed method is evaluated on an augmented dataset containing 3000 image–watershed mask pairs. It was trained on 2400 training images and tested on 600 testing images. The experimental results exemplified significant improvements in segmentation performance, achieving $98.45\%$ validation accuracy, $97.88\%$ validation Dice coefficient, and $95.87\%$ validation intersection over union (IoU) metric scores. These results are superior compared to other conventional and state-of-the-art U-NeT models, thus highlighting the impact of transformer-based hybrid architectures in improving medical image segmentation tasks. Full article

Aiming to address the problem of extracting the remote sensing FTIR spectral characteristics of the hot jet of a certain type of aero-engine under different working conditions, this paper proposes a feature construction algorithm for the remote sensing FTIR spectral characteristics of the aero-engine hot jet based on the fusion of the original spectral features and the deep spectral features. The infrared spectrum was collected at a distance of 280 m, covering the spectral range of 2.5–15 μm with a resolution of 1 cm⁻¹. The Neighborhood–Autoencoder Integration Dual-Branch Network (NAIDN) feature construction algorithm is proposed. This algorithm contains a neighborhood integration branch and an autoencoder branch. The neighborhood integration branch converts the radiation intensity values of discrete wavenumber points into local energy aggregation features through a sliding window, accurately extracting the key physical information in the original spectrum. The autoencoder branch uses a three-layer fully connected neural network architecture to mine the deep spectral features of the spectral data. The algorithms of the two branches not only retain the physical interpretability of spectral analysis but also capture the multi-parameter coupling information hidden in the hot jet spectrum through the representation learning ability of the autoencoder, achieving feature fusion across spatial dimensions. Compared with traditional feature construction algorithms, the dual-branch feature construction algorithm proposed in this paper has stronger comprehensive representation capabilities. The content of carbon dioxide (CO₂) and cyanide groups (-C≡N) in the hot jet under different operating conditions varies significantly. In the experiment, an unsupervised clustering algorithm, the Agglomerative Clustering classifier, is selected, and the classification accuracy of the features extracted by the algorithm in this paper reaches 92.97% on this classifier, thereby verifying the effectiveness of the algorithm in this paper. Full article

Bacterial cellulose (BC) has emerged as a structurally robust, biologically compatible, and highly adaptable biomaterial with significant potential for next-generation wound-care technologies. Its nanofibrillar, extracellular-matrix-like architecture provides exceptional moisture retention, mechanical stability, and conformability, enabling BC to function as an active scaffold rather than a traditional dressing. Advances in chemical modification, composite engineering, and bioactive functionalization, including antimicrobial metals, chitosan, biosurfactants, enzymes, and growth factors, have expanded BC’s therapeutic capabilities. Emerging smart BC dressings integrate biosensors, stimuli-responsive drug release, and 3D-printed architectures tailored to patient-specific wound geometries. Parallel developments in artificial intelligence (AI) are transforming BC production by optimizing bioprocessing, guiding genetic engineering, reducing culture media costs, and enabling real-time quality control, thereby improving scalability and industrial feasibility. These combined innovations position BC as a multifunctional, immunologically instructive, and digitally integrated platform for advanced regenerative wound care. This review reframes BC within the contemporary pathophysiology of chronic wounds, emphasizing its roles in immunomodulation, macrophage polarization, angiogenesis, mechanotransduction, and the disruption of mixed bacterial–fungal biofilms that characterize diabetic foot ulcers and other non-healing wounds. BC hydrogels typically contain >90–99% water and exhibit tensile strengths exceeding 200 MPa, enabling robust mechanical performance in wound environments. Advances in BC composites have demonstrated antimicrobial reductions of 3–5 log units against common chronic-wound pathogens. Full article

Cerium dioxide (CeO₂) has emerged as a structurally versatile oxide for dye-wastewater purification because its architecture, porosity, and surface accessibility can be tuned over a wide range while maintaining good chemical stability and environmental compatibility. Recent studies show that template-free or low-template routes can generate porous, mesoporous, multilayered, and flower-like CeO₂ architectures with rapid dye uptake and, in some systems, adsorption-assisted photocatalytic removal. However, CeO₂-based dye removal has often been discussed either within broad surveys of environmental applications or from composition-centered viewpoints, whereas the more fundamental question is how synthesis route controls architecture formation and how architecture, in turn, governs adsorption and subsequent removal behavior. This mini-review addresses that question from a morphology-centered perspective. It first examines template-free and low-template routes for constructing structured CeO₂, then discusses how porosity, hierarchical assembly, and surface accessibility regulate adsorption kinetics and equilibrium capacity in dye-containing aqueous systems. It further considers adsorption-assisted photocatalytic removal and argues that dark adsorption should be regarded as the structural first step rather than a secondary contribution. On this basis, the review shows that rare-earth doping in these systems is most usefully understood as a secondary tuning strategy that refines an already favorable host architecture by modifying surface interaction, optical response, or reactive-species generation. Overall, the available evidence indicates that CeO₂-based dye-wastewater purification is most meaningfully interpreted through a route–architecture–function framework in which morphology defines the host, adsorption organizes the local reaction environment, and doping serves mainly as structure-assisted tuning. This perspective shifts the design logic of CeO₂ from empirical performance optimization toward rational structure-directed construction of integrated removal platforms. Full article

To investigate the potential of halogen-containing building blocks in the development of artificial carbohydrate receptors, the 1,3,5-trisubstituted 2,4,6-triethylbenzene scaffold with halogenated subunits and classical hydrogen bonding sites was used as a model system. In the first studies, the influence of the presence of halogens on the binding properties of compounds bearing benzamidomethyl units was investigated, whereby the type of halogen and its ring position were varied. The question was whether the presence of halogens could lead to an increase in binding effectivity and whether this increase can be attributed to the formation of halogen bonds (especially for X = Br and I in ortho position) with the sugar substrate or to other effects. The binding studies revealed some interesting relationships between structure and binding affinity for the tested compounds 1–9. For those bearing the halogen substituent in the ortho position to the amide functionality, the binding affinity increases in the expected order 4 (o-F) < 3 (o-Cl) < 2 (o-Br) < 1 (o-I). In the presence of small amounts of water in CDCl₃, an increase in binding strength was observed in comparison to experiments conducted in dry CDCl₃. The present studies aim to provide impulses for the use of halogenated building blocks in the design of artificial carbohydrate receptors. Optimizing the type of halogenated units and the receptor architecture should result in more effective carbohydrate receptors capable of functioning effectively in aqueous media through a combination of different noncovalent interactions. Full article

The Associated Press (AP) Stylebook, endearingly called “the journalist’s bible,” contains thousands of entries outlining style rules and situational guidance. Designed initially for practitioners, the AP stylebook is a seminal resource at many journalism education programs. Its density and complexity as a learning material inherently poses cognitive load risks for novices—and yet—it remains notably under researched. This explanatory sequential mixed methods study explored journalism instructor axiology, pedagogy, and perceptions of teaching effectiveness when introducing AP Style to novice students. Findings revealed that while AP Style remains a pillar of U.S. journalism curriculum, experienced instructors sometimes feel uncertain about the effectiveness of their introductory pedagogy. They described a hodgepodge of methods and design constraints often incongruous with knowledge of human cognitive architecture. We problematize these findings through cognitive load research, recommend Cognitive Apprenticeship Model principles to reduce load-inducing strategies, and suggest directions for future research. Full article

Fibromyalgia (FM) is a complex chronic syndrome marked by widespread musculoskeletal pain, neurocognitive dysfunction (“fibro-fog”), and autonomic disturbances. Clinical management remains challenging due to subjective symptom reporting and the lack of definitive diagnostics. Emerging evidence points to a multifactorial origin involving central sensitization, neuroendocrine imbalance, and systemic immune-inflammatory alterations. A wide array of candidate biomarkers has been reported in FM, encompassing neurotransmitters (serotonin, norepinephrine), excitatory and inhibitory amino acids, metabolic and glycolytic enzymes, stress-related proteins, autoantibodies, oxidative stress markers and pro-inflammatory cytokines. This molecular heterogeneity reflects the systemic and multidimensional nature of FM. However, most of these biomarkers have been primarily investigated in serum or plasma, where analytical validation and reference ranges are more established. In contrast, the exploration of salivary biomarkers—although highly attractive due to its non-invasive, stress-free, and repeatable collection—remains comparatively limited. Saliva contains a reduced concentration range of many systemic markers and is strongly influenced by circadian rhythms, stress, flow rate, and oral health conditions. While promising candidates such as α-amylase, cortisol, calgranulins, and selected metabolic enzymes have shown potential in saliva, many proposed FM-related biomarkers lack full analytical validation, standardized protocols, and clinically defined reference intervals in this matrix. In this context, non-invasive electrochemical biosensors represent a transformative technological approach. Advanced electrode architectures incorporating nucleic acid probes, redox reporters, and nanostructured materials offer high sensitivity in low-volume and low-concentration biofluids such as saliva. The integration of multiplexed biomarker panels into portable platforms could enable real-time, longitudinal monitoring of FM pathophysiology, supporting phenotype stratification, personalized therapeutic adjustment, and objective disease activity tracking. Full article

Angiogenesis is a spatially regulated hallmark of colorectal cancer (CRC) progression, yet current analytical frameworks fail to resolve how nutraceutical bioactive compounds interact with angiogenic signalling within the heterogeneous tumour microenvironment. This review advances a central hypothesis: that the spatial localisation of palm oil mill effluent (POME)-derived bioactive compounds within CRC tumour tissues is predictive of their functional anti-angiogenic activity. POME—the largest waste stream of palm oil processing—contains a chemically diverse array of bioactives, including tocotrienols, phenolics, carotenoids, and fatty acids, with reported antioxidant, anti-inflammatory, and anti-angiogenic properties. However, the existing evidence is predominantly derived from bulk in vitro analyses, limiting mechanistic conclusions about compound behaviour within spatially organised tumour architectures. To address this gap, we propose an integrated framework positioning mass spectrometry imaging (MSI)—across matrix-assisted laser desorption/ionisation (MALDI), desorption electrospray ionisation (DESI), and secondary ion mass spectrometry (SIMS) platforms—as the analytical bridge between compound localisation and angiogenic function. By enabling the label-free, spatially resolved co-localisation of POME-derived compounds with key angiogenic mediators, including VEGF, HIF-1α, and NF-κB, within intact CRC tissues, MSI provides a mechanistic platform that transcends the limitations of conventional molecular analyses. A four-component translational roadmap is outlined, encompassing POME bioactive profiling, spatial compound mapping, angiogenic co-localisation analysis, and functional validation. Critically, the existing evidence on oil palm-derived bioactives is appraised with respect to study quality, mechanistic depth, and translational limitations, identifying the most analytically tractable candidate compounds for spatial investigation. Collectively, this framework positions POME valorisation within a precision nutraceutical oncology paradigm, offering a spatially informed strategy for anti-angiogenic intervention in CRC while simultaneously addressing the environmental burden of palm oil processing waste. Full article

Pig behavior statistics can reflect their health status. Conventional approaches depend on manual observation to derive behavioral information from video recordings, a process that demands substantial time and human effort. To overcome these limitations in indoor intensive farming environments, this study introduces an effective approach for recognizing pig behaviors, employing an enhanced YOLOv8n architecture. The approach utilizes advanced object detection algorithms to automatically identify pig behaviors, including stand, lie, eat, fight, and tail-bite, from overhead video footage of the enclosure. First, images of daily pig behaviors are collected using cameras to build a pig behavior dataset. To boost detection accuracy, the SE attention mechanism is embedded within the feature extraction backbone of the YOLOv8n network to enhance its representational capacity, strengthening the model’s capacity to grasp overarching contextual information and improve the expressiveness of extracted features. The GIoU loss function is employed during training to reduce computational cost and accelerate model convergence. Moreover, integrating Ghost convolution into the backbone significantly reduces both computational complexity and the total number of parameters. The experimental findings reveal that the optimized YOLOv8n model contains just 1.71 million parameters, marking a 42.93% reduction relative to the baseline model. Its floating-point operations total 5.0 billion, indicating a 38.27% decrease, while the mean average precision (mAP@50) reaches 96.8%, surpassing the original by 2.6 percentage points. Compared with other widely used YOLO-based object detection frameworks, the proposed approach achieves notably higher accuracy while requiring significantly lower computational resources and model complexity. Full article