Search Results (4,764)

Analyzing music similarity in large catalogs is challenging because people perceive music differently and important details are found in audio, text, and metadata. This article introduces a multimodal framework that uses an ontology to make music similarity and recommendation more explainable. The framework brings together learned features from audio, lyrics, and other text with structured metadata in a shared similarity space, and then improves ranking with a music ontology that captures relationships between songs, artists, genres, and moods. The design works with any encoder that creates fixed-size features. This study uses strong neural audio and text encoders, mainly based on transformers. This approach allows the system to handle different input types while staying reliable across datasets. This study tests the framework on several open music and audio datasets using content-based retrieval tasks and standard ranking measures. In addition to Configurations C1–C4, this study includes an external content-based reference baseline based on conventional MIR audio descriptors. This baseline represents a signal-level retrieval approach that models complementary aspects of the audio signal, such as timbre, harmony, and spectral characteristics, and is evaluated under the same retrieval protocol as the main framework. It is included to provide an external comparison point outside the proposed C1–C4 design. Compared to audio-only and non-ontological variants within the same framework, the proposed multimodal and ontology-guided configurations achieve better precision, recall, and mean average precision, and also cover more rare content. Visualizations and case studies show that combining different data types and using ontology-based reranking can improve performance and make results easier to interpret. This work lays the groundwork for explainable, cognitively informed music recommendation systems and points to future work in modeling user behavior over time and adapting to different cultures. Full article

Here, we propose a composite system that uses text summarization (TS) and question answering (QA) to supplement the IR process of long documents. Most previous studies have used separate approaches, i.e., either TS or QA. The aim of this paper is to develop an interaction between TS and QA in three stages to enhance IR performance. First, SBERT is used for summarization. Second, an RAG method is employed to retrieve information and generate answers. In the architecture of RAG, retrieval of the document is fulfilled via all-MiniLM-L6-v2, while answer generation is performed via the T5 and BART-large-cnn models. Third, the retrieved answers are assessed and compared with a baseline system in which the documents are treated without summarization. The proposed system aims to improve the quality of retrieved information and accuracy of answers generated by TSQA in a unified pipeline. Experimental evaluation conducted on the NIPS dataset demonstrates that the proposed approach significantly enhances summary informativeness and answer accuracy compared with traditional single-task approaches. The simulation results show improvements of 20.83% in text similarity and 2.38% in BERT scores for answer generation compared with the standard RAG baseline without summarization. Full article

Background/Objectives: Superficial oral mucosal (SOM) lesions are prevalent among patients with Sjögren’s disease (SjD) due to mucosal dryness. Given the limited evidence on screening and referral for SOMs, and the presence of relevant information only in dental clinical notes, a natural language processing (NLP) pipeline was developed to screen for SOMs among SjD patients. This retrospective study analyzed dental clinical notes from 180 linked electronic dental and health records, including both with and without a diagnosis of SjD. Materials and Methods: An annotation schema with four classes (SOMs, signs and symptoms of dry mouth, treatment for xerostomia, referral to specialists) was inductively created using the Extensible Human Oracle Suite of Tools (eHOST) to manually annotate clinical notes. Relevant keyterms were retrieved using a rule-based approach with Python’s Natural Language Toolkit (NLTK). SjD and control groups were compared using Fisher’s Exact tests. Four annotators reviewed ninety-three records. Results: SjD patients (mean age 54.8 ± 11.7 years) had fewer total visits across 15 years but had more dental visits per year (10.2 ± 13.3) than controls. SjD patients were more likely to have oral candidiasis (p = 0.041), exhibit signs and symptoms of dry mouth (p = 0.004), receive treatments for xerostomia (p < 0.001), be treated with cholinergic agonists (p = 0.005), and be referred to a specialist (p = 0.046), but findings were not significant for all SOMs. Additionally, SjD patients had a higher proportion of sialadenitis (p = 0.045), rheumatoid arthritis (p = 0.001), systemic lupus erythematosus (p < 0.001), myalgia/myositis/fibromyalgia (p = 0.010), and anxiety/nervousness (p = 0.004). Conclusions: These findings encourage the feasibility of using text mining from dental clinical notes for screening and management of oral conditions. Full article

Contemporary Music Information Retrieval (MIR) and Natural Language Processing (NLP) systems are increasingly applied to diverse musical traditions, yet they are largely grounded in Western musical and linguistic assumptions. This study examines whether commonly used MIR features and multilingual NLP models adequately represent the acoustic, linguistic, and cultural structures of Maskandi music in comparison to Western music and identifies where representational gaps and biases arise. A multidimensional framework was employed, comprising acoustic and structural MIR analysis, linguistic and semantic lyrical analysis, and bias analysis. A curated dataset of 60 recordings and corresponding lyrics was analysed using rhythm and beat features, pitch contour measures, structural self-similarity, timbre embeddings, semantic similarity, lexical diversity, metaphor density, topic modelling, multilingual embeddings, and dataset-level audits. The results reveal systematic representational failures: beat tracking showed lower median IOI coefficient of variation for Maskandi (0.028) versus Western music (0.040, p = 0.0199) yet exhibited greater algorithmic instability, tempo averaged 131.16 BPM versus 111.69 BPM (p = 0.000262), pitch glide proportions were significantly higher in Maskandi (0.34 vs. 0.16), on-beat energy ratios differed substantially (2.26 vs. 1.19, p < 0.0000007), semantic similarity revealed high intra-genre coherence for Maskandi (0.73) versus Western (0.25), metaphor density approached zero in Maskandi versus up to 7 per 100 words in Western lyrics, topic modeling produced two compact clusters for Maskandi versus 6 dispersed clusters for Western, timbre embeddings achieved a 0.405 silhouette score, dataset audits revealed 0% Maskandi representation across seven major MIR corpora with African traditions comprising <3%. The study concludes that statistical separability does not imply representational adequacy and highlights the need for culturally grounded MIR and NLP representations to support diverse musical traditions. Full article

Background: Generative artificial intelligence (GenAI) models are increasingly used for medical information retrieval, due to their accessibility and efficiency. However, the accuracy and readability of their responses, specifically for upper gastrointestinal cancers, remain inadequately evaluated. This gap highlights the need for rigorous assessment to ensure reliable patient education and clinical integration. Objective: This study aimed to assess the accuracy and readability of responses generated by four prominent GenAI models (Kimi, DeepSeek, ChatGPT, and Gemini) when addressing patient-focused questions related to esophageal and gastric cancers. Methods: Twenty-five standardized medical questions about esophageal and gastric cancer covering domains of disease definition, treatment and management were posed to each model. Responses were assessed by four oncologists for accuracy by a 5-point Likert scale and analyzed for readability using Flesch–Kincaid Reading Ease, Flesch–Kincaid Grade Level, and SMOG metrics. High-interest questions for patients were identified via questionnaires. Results: Comparing the accuracy of GenAI-generated responses, DeepSeek achieved the highest overall accuracy score and outperformed other models in questions about definitions and treatments, while ChatGPT excelled in management-related inquiries. In subgroup analysis, GenAI models exhibited higher accuracy in answering definition and management questions, which patients preferred to inquire, compared with questions about cancer therapies. The responses produced by all models required a reading capacity from 11th-grade to college level. Conclusions: This study revealed that in this comparative evaluation application of GenAI models, DeepSeek provides the most accurate responses for upper GI cancer inquiries about definition and treatment, while ChatGPT showed superiority in management-related questions. However, all models generate texts requiring advanced reading levels, highlighting a need for readability optimization without compromising accuracy. GenAI shows promise for patient education but requires rigorous validation for clinical integration. Full article

Precipitation Use Efficiency (PUE) is a key ecological indicator for evaluating how vegetation converts precipitation into biomass or productivity. A thorough analysis of its quantification methods and driving mechanisms is of great significance for improving regional precipitation use efficiency and ensuring agricultural and ecological water security. In this study, we conducted a comprehensive literature search without time restrictions in the Web of Science and China National Knowledge Infrastructure (CNKI) databases, using “Precipitation Use Efficiency” and “PUE” as core keywords. After retrieval, a strict “independent dual-screening plus cross-checking” procedure was adopted with unified inclusion and exclusion criteria to ensure literature quality. Only highly relevant and methodologically rigorous studies were retained, resulting in a final set of 80 eligible publications. Key information was systematically extracted using content analysis, followed by integrated summarization and inductive analysis. This paper systematically illustrates the ecological connotation of PUE, compares diverse quantification and research methods with their applicable conditions, analyzes spatiotemporal differentiation characteristics and multidimensional driving mechanisms, summarizes practical approaches for PUE improvement, and reviews current research limitations. It represents a systematic integration and refinement of the research framework of precipitation use efficiency. The results can provide targeted theoretical support for revealing the driving mechanisms of PUE and promoting the efficient utilization of precipitation resources. Full article

Soil moisture content (SMC) is a critical factor in agricultural management; however, traditional monitoring methods face limitations regarding spatial resolution and the acquisition of regional dynamics. Unmanned Aerial Vehicle (UAV) remote sensing offers new opportunities for precision monitoring. This study proposes a UAV-based multi-modal remote sensing method for soil moisture estimation. Specifically, novel dual-band and three-band hyperspectral (HS) indices were constructed, and visible (RGB) and thermal infrared (TIR) information were integrated to form a multi-modal data system; simultaneously, multi-modal estimation models were developed by combining four AutoML methods: TPOT, AutoGluon, H2O AutoML, and FLAML. The results indicate that the H2O AutoML model, fusing multi-modal data, exhibited the best performance in estimating soil moisture at depths of 0–20 cm and 20–40 cm (R ≥ 0.72, RMSE 1.99–2.17%), demonstrating superior stability and generalization capabilities compared to other models. This study has made progress in hyperspectral index construction, multi-modal fusion, and soil moisture retrieval, providing a new technical approach for the refined management of agricultural water resources. Full article

As global urbanization accelerates, biophilic urbanism has emerged as a key nature-based strategy for enhancing public health. While plants are critical active agents for psychological restoration, the specific pathways through which vegetation characteristics influence human–environment interactions remain fragmented. This knowledge gap hinders the evidence-based translation of biophilic principles into actionable urban design and governance. This study conducts a systematic bibliometric analysis of 443 peer-reviewed articles (2000–2025) at the intersection of restorative landscapes, urban settings, and plant-based interventions retrieved from the Web of Science Core Collection. Employing multiple visualization tools (VOSviewer, bibliometrix, and CiteSpace), we map publication trends, international collaborations, and thematic evolution. The results demonstrate a significant shift in the field, moving beyond the validation of foundational restorative theories (e.g., ART and SRT) to a more precise, implementation-oriented framework. This shift is characterized by the operationalization of vegetation attributes as controllable design variables, increasingly relating biophilic principles to broader nature-based solutions (NbS) agendas and evidence-informed urban governance. Thematic clustering analysis identified three core knowledge domains: (1) the role of plants as active exposure agents and behavioral mediators in psychological restoration; (2) the impact of specific plant characteristics—such as canopy structure, species diversity, and seasonal variation—on therapeutic outcomes; and (3) the integration of urban green spaces into broader governance frameworks to promote health equity and inclusive well-being. Our analysis highlights that plant-based interventions are evolving from aesthetic ornaments into precision design levers for fostering human–nature interactions. This study provides a science-based foundation for developing practical design guidelines and policy frameworks, shifting biophilic urbanism toward a robust governance strategy for creating equitable, restorative, and resilient cities. Full article

Unmanned aerial vehicle (UAV)-based hyperspectral imaging (HSI) has been increasingly utilized for fine-scale surface characterization and quantitative retrieval due to its capability of capturing dense spectral information at ultra-high spatial resolution. However, UAV-HSI analysis remains challenging due to high dimensionality, noise and within-class variability, as well as limited cross-flight consistency under varying acquisition conditions. Deep learning (DL) has therefore attracted growing attention by enabling spectral-spatial representation learning and more robust inference under residual degradations and domain shifts. This review summarizes DL approaches for UAV-HSI analytics and organizes the literature along a complete workflow, from imaging principles, preprocessing, and correction to DL architectures, core tasks, and representative applications, to provide guidance for future research and applications. The reviewed papers demonstrate that DL exhibits great potential and a promising future in UAV-HSI analysis. Full article

Over six decades, satellite remote sensing of water resources has evolved from manual interpretation of weather photographs to AI systems that learn hydrologic predictions directly from satellite imagery. This review traces that evolution through the observation-to-inference arc—a framework for the progressively tightening coupling between what satellites observe and what hydrologists infer. Using illustrative applications in precipitation, evapotranspiration, soil moisture, snow, surface water, and groundwater, we show how early observations (1960–1985) remained disconnected from operational hydrology; how calibrated retrieval algorithms (1985–2000) established a one-way pipeline from satellites to models; how operational infrastructure (2000–2015), anchored by MODIS, GRACE, GPM, and Sentinel, achieved assimilative coupling through computational feedback between models and observations; and how deep learning (2015–present) is beginning to collapse this pipeline. Multi-source data fusion has been a recurring enabler at each stage. We articulate a four-level AI vision and research trajectory, from AI-assisted interpretation through AI-native retrieval and AI-driven inference to autonomous Earth observation intelligence. Persistent challenges in mission continuity, calibration, equity of access, and translating satellite-derived information into operational water management decisions provide essential context for evaluating both the promise and limits of this trajectory. Full article

Photoplethysmography (PPG) has emerged as an attractive modality for non-invasive cardiovascular monitoring due to its low cost, unobtrusive nature, and ubiquity in consumer wearable devices. Despite its potential, existing PPG-based arrhythmia detection systems remain limited in scope: (i) most target only atrial fibrillation, (ii) temporal localization of abnormal segments is rarely provided, and (iii) deep learning models lack explainability, hindering adoption in clinical workflows. We present a comprehensive and fully integrated framework for multi-class arrhythmia detection, segmentation, and explainability based on PPG waveforms, Heart Rate Variability (HRV), and structured clinical metadata. The proposed system introduces a CLIP-style contrastive learning module aligning PPG waveforms with clinical variables and rhythm-state textual descriptions using BioBERT; a multitask U-Net architecture performing 4-class classification and 1D segmentation; a Retrieval-Augmented Generation (RAG) pipeline leveraging Gemini Flash large language models to produce guideline-grounded diagnostic reports; and a real-time Streamlit-based web platform supporting inference, visualization, and database storage. The system significantly improves classification accuracy (from 86.27% to 91.19%) and segmentation Dice (from 0.5815 to 0.7167). These results demonstrate the feasibility of a robust, multimodal, and explainable PPG-based arrhythmia monitoring system for real-world applications. Full article

Hyperspectral remote sensing provides essential biochemical and structural information for crop disease monitoring, yet its application to cotton Verticillium wilt has largely focused on single-period evaluations or multi-temporal classifications. Such approaches overlook the progressive nature of this vascular disease, whose pigment, water, and mesophyll responses evolve over time, making temporal hyperspectral information critical for reliable severity estimation but still insufficiently utilized. To overcome this limitation, we conducted daily time-series observations on cotton leaves and collected 2895 hyperspectral reflectance measurements and 770 high-resolution RGB images together with disease severity records, generating a temporally dense spectral-severity dataset spanning symptom-free to severe stages. Five categories of disease-related vegetation indices were derived and organized into 5-day spectral–temporal slices. Based on these features, we introduce a dual-branch Transformer-TCN model that integrates global temporal dependencies captured by self-attention with local temporal variations resolved by dilated causal convolutions for severity inversion. The model delivers the strongest performance with an R² of 0.8813, exceeding multiple single and hybrid time-series alternatives by 0.0446–0.1407 in R², equivalent to a relative improvement of 5.33–19.00%. Temporal spectral features also outperform their non-temporal counterparts, highlighting that disease progression dynamics captured by time-series spectra are critical for reliable severity retrieval. Feature contribution analysis indicates that the blue red index BRI provides the highest contribution, consistent with the single-index time-series modelling results. Photosynthesis- and water-related indices provide secondary but complementary support. Collectively, our results demonstrate that the dual-branch Transformer-TCN model can capture complex spectral–temporal relationships between cotton Verticillium wilt and disease severity, providing methodological support for crop disease monitoring and evaluation. Full article

Pedestrian gender recognition plays an important role in pedestrian analysis and intelligent video applications, for example, in demographic statistics, soft biometric analysis, and context-aware person retrieval. However, it remains a challenging task owing to viewpoint variations, illumination changes, occlusions, and low image quality in real-world imagery. To address these issues, an effective adaptive decision fusion framework, termed the Decision Fusion Learning Network (DFLN), is proposed in this paper. The key novel aspect of DFLN is that it effectively explores both an appearance-centered view that emphasizes detailed texture and clothing information and a structure-centered view that captures rich contour and structural information for pedestrian gender recognition. To realize DFLN, a Parallel CNN Prediction Probability Learning Module (PCNNM) is first constructed to independently learn modality-specific probabilities from color image and edge maps. Subsequently, a learnable Decision Fusion Module (DFM) is designed to fuse the modality-specific probabilities and explore their complementary merits for realizing accurate pedestrian gender recognition. The DFM can be easily coupled with the PCNNM, forming an end-to-end decision fusion learning framework that simultaneously learns the feature representations and carries out adaptive decision fusion. Experiments on two pedestrian benchmark datasets, named PETA and PA-100K, show that DFLN achieves competitive or superior performance compared with several state-of-the-art pedestrian gender recognition methods. Extensive experimental analysis further confirms the effectiveness of the proposed decision fusion strategy and its favorable generalization ability under domain shift. Full article

Data centers are critical digital infrastructure supporting cloud computing, artificial intelligence, and global information services. Despite their high-reliability design, they remain vulnerable to fire incidents due to continuous operation, high electrical loads, dense power systems, and the increasing use of lithium-ion batteries. Although such events are rare, their consequences can be severe, including service disruption, equipment damage, financial loss, and risks to data integrity. This study presents a systematic literature review of fire safety risk management frameworks in data centers, following PRISMA guidelines. Peer-reviewed studies published between 2020 and 2025 were retrieved from Scopus and Web of Science, screened, and appraised using structured quality criteria. Twelve empirical studies were synthesized and benchmarked against NFPA 75 and NFPA 76 standards. The findings are organized into three domains: Strategic Management, Fire Risk, and Fire Preparedness. The results show a strong focus on technical prevention and electrical hazards, while organizational readiness, emergency response, and recovery remain underexplored. Benchmarking indicates that industry standards adopt a more comprehensive lifecycle approach than the academic literature. This study reframes data center fire safety as a socio-technical reliability system and highlights critical gaps, providing a foundation for future research and improved fire safety governance and resilience. Full article

With the increasing penetration of converter-based devices, harmonic distortion has become a major challenge for power quality monitoring in large-scale power systems. This study presents a systematic review of methods for modeling harmonic sources and their applicability to real-time monitoring of power distribution systems. The review was conducted following PRISMA guidelines, considering literature published between 2000 and 2026. Searches were performed across Scopus, IEEE Xplore, Web of Science, ScienceDirect, and MDPI using predefined keywords. A total of 128 peer-reviewed journal articles were included. Potential sources of bias were qualitatively assessed, including selection, retrieval, and classification bias; however, residual bias may still arise from database selection, keyword design, and study classification. A structured comparative framework is introduced, based on a six-dimension coverage scoring scheme and maturity analysis, enabling consistent evaluation across both methodological and deployment aspects. The robustness of this framework was evaluated using leave-one-out and perturbation analyses, indicating low variability in coverage scores and stable rankings across both corpora. A taxonomy of harmonic source modeling approaches is proposed. Comparative synthesis indicates that measurement-based approaches, particularly those leveraging distribution-level PMUs, show strong potential for real-time monitoring. Key challenges include D-PMU placement, data integration, and computational scalability. Future work should focus on physics-informed AI and digital twin-based monitoring. Full article