Search Results (64)

This brief survey acts as a fundamental resource for researchers beginning their exploration into fake news detection. It emphasizes the importance of dataset quality and diversity in enhancing the effectiveness of detection models, detailing key features, labeling systems, and prevalent biases. It also presents the challenges and limitations. By addressing ethical considerations (such as privacy and consent, societal impacts, transparency, and accountability) and best practices (annotation methodologies, real-world dynamics, reliability, and validity), we offer a thorough overview of current datasets. Additionally, our contribution includes a GitHub repository that aggregates publicly available datasets into a single, easily accessible portal, thereby supporting further research and development in the fight against fake news. Full article

Background: The traditional formula Liqi Yangyin (LQYY) has shown clinical and preclinical efficacy for depression with constipation, yet its molecular mechanisms remain incompletely defined. This study aimed to elucidate its mechanisms using an integrative approach. Methods: Constituents of LQYY were profiled by UPLC-MS/MS and integrated with network pharmacology and molecular docking to identify brain-accessible components and putative targets. A chronic unpredictable mild stress (CUMS) model was used for experimental validation. Outcomes included behavioral tests (sucrose preference test, open field test, and forced swimming test), gastrointestinal indices, including fecal water content, time of first black stool, and intestinal propulsion rate, histopathology of the prefrontal cortex (PFC) and colon, TUNEL staining, NeuN immunofluorescence, Western blotting, and qRT-PCR. Results: LQYY attenuated CUMS-induced weight loss and depressive-like behaviors and improved intestinal transit metrics. It reduced neuronal apoptosis in the PFC and ameliorated colonic injury. Mechanistically, docking and enrichment analyses highlighted hub targets (STAT3, AKT1, ESR1, IL-6, TNF, TP53) and the JAK/STAT pathway. In vivo, LQYY decreased IL-6, TNF-α, ESR1, TP53, and STAT3, and increased AKT1 in the PFC and colon; it also reduced the TUNEL-positive rate and restored NeuN labeling, upregulated Bcl-2, and downregulated p-JAK2/JAK2 and p-STAT3/STAT3 ratios, and the expression of Bax and cleaved-caspase-3 in the PFC, consistent with the suppression of pro-inflammatory and apoptotic signaling. Conclusions: LQYY exerts antidepressant and pro-motility effects in CUMS mice by modulating JAK2/STAT3-centered networks and inhibiting neuronal apoptosis, thus supporting a multi-component, multi-target strategy for treating depression with constipation, and providing a defined molecular hypothesis for future investigation. Full article

Green fintech merges sustainable finance with data-intensive innovation, but national translations of EU rules can create regulatory risk. This study examines how such risk manifests in Central Europe and which policy tools mitigate it. We develop a three-dimension framework—regulatory clarity and scope, supervisory consistency, and innovation facilitation—and apply a comparative qualitative design to Hungary, Slovakia, Czechia, and Poland. Using a common EU baseline, we compile coded national snapshots from primary legal texts, supervisory documents, and recent scholarship. Results show material cross-country variation in labelling practice, soft-law use, and testing infrastructure: Hungary combines central-bank green programmes with an innovation hub/sandbox; Slovakia aligns with ESMA and runs hub/sandbox, though the green-fintech pipeline is nascent; Czechia applies a principles-based safe harbour and lacks a national sandbox; and Poland relies on a virtual sandbox and binding interpretations with limited soft law. These choices shape approval timelines, retail penetration, and cross-border portability of green-labelled products. We conclude with a policy toolkit: labelling convergence or explicit safe harbours, a cross-border sandbox federation, ESRS/ESAP-ready proportionate disclosures, consolidation of recurring interpretations into soft law, investment in suptech for green-claims analytics, and inclusion metrics in sandbox selection. Full article

Financial fraud detection is critical to modern economic security, yet remains challenging due to collusive group behavior, temporal drift, and severe class imbalance. Most existing graph neural network (GNN) detectors rely on pairwise edges and correlation-driven learning, which limits their ability to represent high-order group interactions and makes them vulnerable to spurious environmental cues (e.g., hubs or temporal bursts) that correlate with labels but are not necessarily causal. We propose Causal-DHG, a dynamic hypergraph framework that integrates hypergraph modeling, causal intervention, and multi-view contrastive learning. First, we construct label-agnostic hyperedges from publicly available metadata to capture high-order group structures. Second, a Multi-Head Spatio-Temporal Hypergraph Attention encoder models group-wise dependencies and their temporal evolution. Third, a Causal Disentanglement Module decomposes representations into causal and environment-related factors using HSIC regularization, and a dictionary-based backdoor adjustment approximates the interventional prediction $P(Y\mid do(C\left)\right)$ to suppress spurious correlations. Finally, we employ self-supervised multi-view contrastive learning with mild hypergraph augmentations to leverage unlabeled data and stabilize training. Experiments on YelpChi, Amazon, and DGraph-Fin show consistent gains in AUC/F1 over strong baselines such as CARE-GNN and PC-GNN, together with improved robustness under feature and structural perturbations. Full article

This study aims to characterize tissue-specific expression patterns in Hanwoo steers by identifying co-expression modules, functional pathways, and hub genes related to fat and muscle traits using Weighted Gene Co-expression Network analysis (WGCNA). RNA-Seq data were generated from three muscle tissues (longissimus muscle, tenderloin, and rump) and two fat tissues (back fat and abdominal fat) collected from six 30-month-old Hanwoo steers. Quality control of raw sequencing reads was performed using FastQC, and trimmed reads were aligned to the bovine reference genome (ARS-UCD1.3) using HISAT2. We also identified a gene co-expression network via WGCNA using normalized gene expression values. Modules were defined based on topological overlap and correlated with tissue-specific expression patterns. Modules with a significant association (p < 0.05) were used for functional enrichment based on Gene Ontology (GO) and KEGG pathways, as well as Protein–Protein Interaction Network analysis. A total of seven co-expression modules were identified by WGCNA and labeled in distinct colors (yellow, blue, red, brown, turquoise, green, black). Among them, the yellow and blue modules were positively associated with back fat, while the turquoise and green modules showed a negative correlation with abdominal fat. Additionally, the turquoise or green module was positively correlated with longissimus and rump tissues, indicating distinct gene expression patterns between fat and muscle. This study identified key co-expression modules and hub genes associated with muscle and fat metabolism. Notably, ARPC5 (blue module) was involved in lipid metabolism and energy storage, whereas AGPAT5 (turquoise module) was linked to maintaining muscle cell structure and function. These findings reveal biological mechanisms for tissue-specific gene regulation, providing targets for enhancing meat quality in Hanwoo. Full article

Depression risk prediction benefits from approaches that go beyond binary labels by offering interpretable, quantitative views of affective states. This study presents a web-based platform that estimates depression risk by combining Korean Valence–Arousal–Dominance (VAD) regression with a structured, multi-anchor distance scoring method. We construct a Korean VAD–labeled resource by integrating the NRC-VAD Lexicon, the AI Hub emotional dialogue corpus, and translated EmoBank entries, and fine-tune a KLUE-RoBERTa regression model to predict sentence-level VAD vectors. Depression risk is then derived as the mean Euclidean distance from the predicted VAD vector to depressive anchor vectors and normalized into an interpretable risk index. In evaluation, the approach shows strong agreement with ground truth (Pearson’s $r=0.87$) and supports accurate risk screening when thresholded. The platform provides intuitive visual feedback for end users and monitoring tools for professionals, highlighting the practicality of integrating interpretable VAD modeling with lightweight scoring in real-world, web-based mental health support. Full article

The increasing demand for accessible and efficient machine learning solutions has led to the development of the Adaptive Learning Framework (ALF) for multi-class, single-label image classification. Unlike existing low-code tools, ALF integrates multiple transfer learning backbones with a guided, adaptive workflow that empowers non-technical users to create custom classification models without specialized expertise. It employs pre-trained models from TensorFlow Hub to significantly reduce computational costs and training times while maintaining high accuracy. The platform’s User Interface (UI), built using Streamlit, enables intuitive operations, such as dataset upload, class definition, and model training, without coding requirements. This research focuses on small-scale image datasets to demonstrate ALF’s accessibility and ease of use. Evaluation metrics highlight the superior performance of transfer learning approaches, with the InceptionV2 model architecture achieving the highest accuracy. By bridging the gap between complex deep learning methods and real-world usability, ALF addresses practical needs across fields like education and industry. Full article

The increasing use of Arabic and Urdu on social media platforms, particularly Twitter, has created a growing need for robust Named Entity Recognition (NER) systems capable of handling noisy, informal, and code-mixed content. However, both languages remain significantly underrepresented in NER research, especially in social media contexts. To address this gap, this study makes four key contributions: (1) We introduced a manual entity consolidation step to enhance the consistency and accuracy of named entity annotations. In the original datasets, entities such as person names and organization names were often split into multiple tokens (e.g., first name and last name labeled separately). We manually refined the annotations to merge these segments into unified entities, ensuring improved coherence for both training and evaluation. (2) We selected two publicly available datasets from GitHub—one in Arabic and one in Urdu—and applied two novel strategies to tackle low-resource challenges: a joint multilingual approach and a translation-based approach. The joint approach involved merging both datasets to create a unified multilingual corpus, while the translation-based approach utilized automatic translation to generate cross-lingual datasets, enhancing linguistic diversity and model generalizability. (3) We presented a comprehensive and reproducible pseudocode-driven framework that integrates translation, manual refinement, dataset merging, preprocessing, and multilingual model fine-tuning. (4) We designed, implemented, and evaluated a customized XLM-RoBERTa model integrated with a novel attention mechanism, specifically optimized for the morphological and syntactic complexities of Arabic and Urdu. Based on the experiments, our proposed model (XLM-RoBERTa) achieves 0.98 accuracy across Arabic, Urdu, and multilingual datasets. While it shows a 7–8% improvement over traditional baselines (RF), it also achieves a 2.08% improvement over a deep learning (BiLSTM = 0.96), highlighting the effectiveness of our cross-lingual, resource-efficient approach for NER in low-resource, code-mixed social media text. Full article

This study aims to investigate the effect of the Pd(II) complex on HeLa cells using computational biology and proteomic analysis. [Pd(dach)Cl₂]-treated HeLa cells were subjected to comparative proteomics analysis using label-free data-independent liquid chromatography-tandem mass spectrometry (LC-MS/MS). In parallel, the informational spectrum method (ISM) was used to predict potential protein interactors of the [Pd(dach)Cl₂] complex in HeLa cells. Proteomics analysis revealed 121 differentially abundant proteins (DAPs). Enrichment analysis of Gene Ontology (GO) annotations revealed ATP hydrolysis and RNA/protein binding as the top molecular functions and RNA splicing and protein–RNA complex organization as the top biological processes. Enrichment analysis of altered canonical pathways pointed out spliceosome and ribosome pathways. The top hub proteins with potential regulatory importance encompassed ribosomal proteins, translational and transcriptional factors, and components of the ribosome assembly machinery. ISM and cross-spectral analysis identified the nucleoplasm and sensor of the single-stranded DNA (SOSS DNA) complex. Proteome analysis showed that [Pd(dach)Cl₂] targets proteins involved in ribosomal biogenesis and RNA splicing, whereas theoretical prediction implies also potential effect on p53 signaling pathway, and thus, alterations of the expression of regulatory proteins involved in cell survival and proliferation. These findings underscore the potential of Pd(II) complexes as anti-cancer agents, warranting further exploration and detailed functional validation. Full article

Digital transformation is increasingly relevant in food certification systems, improving processes, coordination, and data accessibility. In state-led certification systems, public entities hold a political mandate to promote digital transformation, yet little is known about digital maturity in these systems or how to assess it. This study assesses the digital maturity of a state-led food certification system in Baden-Württemberg, Germany, focusing on private sector stakeholders involved in its administration. Additionally, it examines potential measures that the governing public entity can take and evaluates the suitability of the methods used. A total of 25 out of 43 organisations were surveyed using the Digital Maturity Assessment (DMA) framework validated for the European Union (EU). Six dimensions were analysed: Digital Business Strategy, Digital Readiness, Human-Centric Digitalisation, Data Management, Automation and Artificial Intelligence, and Green Digitalisation. Data Management and Human-Centric Digitalisation were the most developed, highlighting strong data governance and workforce engagement. Automation and Artificial Intelligence were ranked lowest, reflecting minimal adoption but also indicating that not all dimensions might be of the same relevance for the variety of organisations. The variability in scores and organisation-specific relevance underscores the European DMA framework’s value, particularly due to its subsequent tailored consultation process and its integration into EU policy. Full article

Pedestrian panic behavior is a primary cause of overcrowding and stampede accidents in public micro-road network areas with high pedestrian density. However, reliably detecting such behaviors remains challenging due to their inherent complexity, variability, and stochastic nature. Current detection models often rely on single-modality features, which limits their effectiveness in complex and dynamic crowd scenarios. To overcome these limitations, this study proposes a contour-driven multimodal framework that first employs a CNN (CDNet) to estimate density maps and, by analyzing steep contour gradients, automatically delineates a candidate panic zone. Within these potential panic zones, pedestrian trajectories are analyzed through LSTM networks to capture irregular movements, such as counterflow and nonlinear wandering behaviors. Concurrently, semantic recognition based on Transformer models is utilized to identify verbal distress cues extracted through Baidu AI’s real-time speech-to-text conversion. The three embeddings are fused through a lightweight attention-enhanced MLP, enabling end-to-end inference at 40 FPS on a single GPU. To evaluate branch robustness under streaming conditions, the UCF Crowd dataset (150 videos without panic labels) is processed frame-by-frame at 25 FPS solely for density assessment, whereas full panic detection is validated on 30 real Itaewon-Stampede videos and 160 SUMO/Unity simulated emergencies that include explicit panic annotations. The proposed system achieves 91.7% accuracy and 88.2% F1 on the Itaewon set, outperforming all single- or dual-modality baselines and offering a deployable solution for proactive crowd safety monitoring in transport hubs, festivals, and other high-risk venues. Full article

Multi-label classification (MLC) plays a crucial role in various real-world scenarios. Prediction with nearest neighbors has achieved competitive performance in MLC. Hubness, a phenomenon in which a few points appear in the k-nearest neighbor (kNN) lists of many points in high-dimensional spaces, may significantly impact machine learning applications and has recently attracted extensive attention. However, it has not been adequately addressed in developing MLC algorithms. To address this issue, we propose a hubness-aware kNN-based MLC algorithm in this paper, named multi-label hubness information-based k-nearest neighbor (MLHiKNN). Specifically, we introduce a fuzzy measure of label relevance and employ a weighted kNN scheme. The hubness information is used to compute each training example’s membership in relevance and irrelevance to each label and calculate weights for the nearest neighbors of a query point. Then, MLHiKNN exploits high-order label correlations by training a logistic regression model for each label using the kNN voting results with respect to all possible labels. Experimental results on 28 benchmark datasets demonstrate that MLHiKNN is competitive among the compared methods, including nine well-established MLC algorithms and three commonly used hubness reduction techniques, in dealing with MLC problems. Full article

Background/Objectives: Most individuals who have a familial or clinical risk of developing psychosis remain free from psychopathology. Identifying neural markers of resilience in these at-risk individuals may help clarify underlying mechanisms and yield novel targets for early intervention. However, in contrast to studies on risk biomarkers, studies on neural markers of resilience to psychosis are scarce. The current study aimed to identify potential brain markers of resilience to psychosis. Methods: A systematic review of the literature yielded a total of 43 MRI studies that reported resilience-associated brain changes in individuals with an elevated risk for psychosis. Label-based meta-analysis was used to synthesize findings across MRI modalities. Results: Resilience-associated brain changes were significantly overreported in the default mode and language network, and among highly connected and central brain regions. Conclusions: These findings suggest that the DMN and language-associated areas and central brain hubs may be hotspots for resilience-associated brain changes. These neural systems are thus of key interest as targets of inquiry and, possibly, intervention in at-risk populations. Full article

In this paper, we define and characterize the embedding of edges and higher-order entities in directed graphs (digraphs) and relate these embeddings to those of nodes. Our edge-centric approach consists of the following: (a) Embedding line digraphs (or their iterated versions); (b) Exploiting the rank properties of these embeddings to show that edge/path similarity can be posed as a linear combination of node similarities; (c) Solving scalability issues through digraph sparsification; (d) Evaluating the performance of these embeddings for classification and clustering. We commence by identifying the motive behind the need for edge-centric approaches. Then we proceed to introduce all the elements of the approach, and finally, we validate it. Our edge-centric embedding entails a top-down mining of links, instead of inferring them from the similarities of node embeddings. This analysis is key to discovering inter-subgraph links that hold the whole graph connected, i.e., central edges. Using directed graphs (digraphs) allows us to cluster edge-like hubs and authorities. In addition, since directed edges inherit their labels from destination (origin) nodes, their embedding provides a proxy representation for node classification and clustering as well. This representation is obtained by embedding the line digraph of the original one. The line digraph provides nice formal properties with respect to the original graph; in particular, it produces more entropic latent spaces. With these properties at hand, we can relate edge embeddings to node embeddings. The main contribution of this paper is to set and prove the linearity theorem, which poses each element of the transition matrix for an edge embedding as a linear combination of the elements of the transition matrix for the node embedding. As a result, the rank preservation property explains why embedding the line digraph and using the labels of the destination nodes provides better classification and clustering performances than embedding the nodes of the original graph. In other words, we do not only facilitate edge mining but enforce node classification and clustering. However, computing the line digraph is challenging, and a sparsification strategy is implemented for the sake of scalability. Our experimental results show that the line digraph representation of the sparsified input graph is quite stable as we increase the sparsification level, and also that it outperforms the original (node-centric) representation. For the sake of simplicity, our theorem relies on node2vec-like (factorization) embeddings. However, we also include several experiments showing how line digraphs may improve the performance of Graph Neural Networks (GNNs), also following the principle of maximum entropy. Full article

In the realm of software development, quality reigns supreme, but the ever-present danger of vulnerabilities threatens to undermine this fundamental principle. Insufficient early vulnerability identification is a key factor in releasing numerous apps with compromised security measures. The most effective solution could be using machine learning models trained on labeled datasets; however, existing datasets still struggle to meet this need fully. Our research constructs a vulnerability dataset for Android application source code, primarily based on the Common Vulnerabilities and Exposures (CVE) system, using data derived from real-world developers’ vulnerability-fixing commits. This dataset was obtained by systematically searching such commits on GitHub using well-designed keywords. This study created the dataset using vulnerable code snippets from 366,231 out of 2.9 million analyzed repositories. All scripts used for data collection, processing, and refinement and the generated dataset are publicly available on GitHub. Experimental results demonstrate that fine-tuned Support Vector Machine and Logistic Regression models trained on this dataset achieve an accuracy of 98.71%, highlighting their effectiveness in vulnerability detection. Full article