With urban development, large buildings with complex structures have increased, creating a growing demand for indoor navigation. Existing research focuses on semantic-augmented localization or spatial relationship descriptions but lacks language-driven frameworks. This paper proposes an indoor semantic knowledge graph and natural language topology framework to address the limitations of physical signal-dependent technologies and non-linguistic topological models. It pioneered an end-to-end navigation paradigm based on natural language semantics, completely eliminating the reliance on Bluetooth/WiFi signals or geometric coordinates. By constructing an indoor semantic knowledge graph and a natural language spatial topology framework, the mapping from natural language descriptions to dynamic navigation instructions is achieved, solving the problems of high hardware cost and poor signal stability of traditional methods. Experiments confirm the framework generates human-understandable instructions, offering a new device-independent navigation paradigm.
Syslog data generated by software systems constitutes a critical source for system anomaly detection. However, existing anomaly detection approaches are highly susceptible to log parsing errors. In addition, sequence models based on a single architecture exhibit limited capacity to capture complex temporal dependencies in log sequences, resulting in suboptimal detection accuracy. In this work, ParseRobustLog is proposed as an efficient log sequence anomaly detection framework based on a Switch network. It incorporates a contrastive learning strategy to align encoded representations with target log template features, thereby mitigating the adverse impact of parsing errors. Furthermore, the Switch network dynamically integrates multiple sequence models and adaptively selects the most suitable model for each log sequence, significantly enhancing detection performance, particularly on datasets with sparse log events. Experiments on multiple widely used datasets demonstrate that ParseRobustLog consistently outperforms existing state-of-the-art anomaly detection methods.
Clickbait headlines undermine trust in digital journalism by exploiting readers' emotions rather than delivering substantive information. Prior research has largely focused on coarse sentiment polarity, leaving the role of fine-grained emotions underexplored. This study proposes a multi-stage artificial intelligence framework that incorporates RoBERTa-based emotion classification, Spearman correlation analysis, topic modeling, and association rule mining to analyze the clickbait data. Findings from the framework reveal that surprise is the strongest positive correlation of clickbait severity, while sadness, anger, and fear are negatively associated. Also, high-severity clickbait relies on isolated surprise, whereas legitimate news often combines surprise with a neutral tone. These results define an emotional fingerprint of clickbait, offering new theoretical insights into emotional persuasion and practical guidance for news recommendation systems and content governance on social platforms.
The modern power grid has evolved into a large-scale, cyber-physical system with complex interactions among heterogeneous entities. Accurate modeling and analysis are essential, yet grid data often suffer from missing values due to sensor failures and communication issues. Tensor factorization is a powerful tool for capturing high-order dependencies and enabling data imputation, but existing methods face scalability and synchronization bottlenecks on large-scale heterogeneous tensors. To address these challenges, this paper proposes a high-performance tensor factorization framework for large-scale power grid data. The approach represents monitoring signals as structured high-order tensors, adopts a 3D block partitioning strategy for efficient distribution, and designs a lock-free scheduling scheme for stochastic updates to minimize synchronization overhead. Complexity analysis confirms the method’s efficiency in both time and memory. Experiments on four real-world power grid datasets demonstrate superior scalability and efficiency performance compared with state-of-the-art methods.
Clickbait headlines undermine trust in digital journalism by exploiting readers’ emotions rather than delivering substantive information. Prior research has largely focused on coarse sentiment polarity, leaving the role of fine-grained emotions underexplored. This study proposes a multi-stage artificial intelligence framework that incorporates RoBERTa-based emotion classification, Spearman correlation analysis, topic modeling, and association rule mining to analyze the clickbait data. Findings from the framework reveal that surprise is the strongest positive correlation of clickbait severity, while sadness, anger, and fear are negatively associated. Also, high-severity clickbait relies on isolated surprise, whereas legitimate news often combines surprise with a neutral tone. These results define an emotional fingerprint of clickbait, offering new theoretical insights into emotional persuasion and practical guidance for news recommendation systems and content governance on social platforms.
Accurate and interpretable prediction of hot rolling force is essential for optimizing material deformation, improving product quality, and reducing energy consumption in the hot rolling process. However, due to idealized assumptions and dynamic process variations, the in-field mechanistic models often lack sufficient accuracy in practical settings, which may lead to issues such as excessive thickness tolerance deviations and plate shape defects. To address this limitation, the authors propose a mechanism–data fusion model based on the LightGBM (LGBM) framework, which incorporates both mechanistic parameters and key process variables as structured input features, aiming to enhance predictive performance. To further optimize the model, Particle Swarm Optimization (PSO) is employed for hyperparameter tuning within the LGBM framework. Comparative experiments demonstrate the proposed model outperforms alternatives on the test set, with 94.05% of predictions having relative errors within ±6% and a coefficient of determination (R2) of 0.977—indicating strong engineering reliability.
This paper presents a novel web service for topic modeling of Classical Chinese poetry. Classical poems are typically short, which causes data sparsity problems for standard topic models, and this leads to poor service output quality. To solve this, the authors design a new service, and at its core is a new algorithm that combines character embeddings and the TextRank algorithm to construct ranked character distributions, which serve as auxiliary semantic information to enrich the input and alleviate data sparsity. They evaluated the service on two large datasets of Tang and Song poetry, and measured service quality through Coherence and PMI scores. Experimental results demonstrate that the proposed service significantly outperforms mainstream baseline methods in the consistency of topic mining, validating its effectiveness as an efficient cultural computation service.
This paper explores the innovative application of convolutional neural networks (CNNs) and human-computer interaction (HCI) in art design, facilitated by advanced web-services technology. The integration of these technologies aims to streamline the creative process, making it more accessible and cost-effective for artists. By leveraging deep learning algorithms for style transfer and visual content processing, this research demonstrates how CNNs can transform user inputs into unique artworks. Furthermore, HCI principles ensure a seamless interface between users and digital tools, enhancing usability and efficiency. This study also addresses the challenges associated with artificial intelligence in art creation, including ethical considerations and the preservation of cultural heritage. Through practical implementations and case studies, the paper highlights the potential for web-based platforms to democratize art creation while maintaining artistic integrity.
Urban flood prevention faces challenges in real-time monitoring, data integration, and decision-making support, particularly under constraints of sensitive data security and efficient emergency response. This study proposes an artificial intelligence-assisted decision-making framework for urban flood prevention, integrating multi-dimensional data fusion via an intranet. A city-level digital twin platform is developed, combining real-time sensor data, meteorological information, and social media data using a geographic information system, the Internet of Things, and artificial intelligence algorithms. The framework ensures secure data processing through intranet-based operations, enhancing flood prediction accuracy and emergency response efficiency. The system was successfully implemented in Hebi City, Henan Province, demonstrating significant improvements in flood risk management. This research advances urban resilience and provides a scientific basis for flood management in smart cities.
Large-scale language models have demonstrated robust language understanding and generation capabilities, enabling them to tackle various complex natural language processing tasks. However, for domain-specific tasks like healthcare that require specialized expertise, relying solely on large language models for dialogue generation is insufficient. Moreover, this paper aims to improve the performance of models in medical conversations and enhance the interpretability of the intermediary processes. It argues that leveraging diverse knowledge and agent-based architecture can significantly address the challenges. We introduce an agent-based adaptive medical dialogue service (AMDS) for personalized healthcare. This service utilizes large language models as its cognitive core and integrates medical knowledge extracted from knowledge graph and process knowledge. Extensive experiments show that AMDS outperforms baselines in multi-turn medical dialogue generation tasks.
Web service clustering technique can effectively improves the service retrieval efficiency. Service networks offer new possibilities to handle the huge growth of Web services; Community detection is one of the important tasks in Web data mining to efficiently analyze and understand the structural properties and group characteristics of various networks. In view of this, this paper proposes a service clustering method based on unsupervised community detection. First, a structure center update strategy is used to overcome the dependence on the initial structure center; Second, the label propagation model is based on the GCN model as a base module, which can utilize both the network topology and node attributes. In order to improve the model's label propagation capability, the method extends the pseudo-label set as supervisory information to train the model and is used to infer the community labels of the remaining nodes. Finally, experiments conducted on 4 real networks show that the method has better community detection performance.
Industrial diagnostic systems play an important role in food manufacturing by ensuring rapid detection of defective components and precise identification of systemic dysfunction. This article proposes a diagnostic model for the pasteurization process to enhance dairy production systems. The authors found that, when a breakdown occurs, the acquisition system stops providing necessary data for diagnostics. To solve this problem, the authors used digital twin (DT) engineering to generate missing values and build a learning model based on reinforcement learning (RL). The effectiveness of this approach was validated through implementation at Aures Batna Dairy, a prominent player in Algeria's dairy industry. Experiments demonstrated the superior efficiency of this method; its precision surpassed that of traditional data imputation techniques by a significant margin.
The integration of artificial intelligence (AI) in education, particularly in engineering drawing courses, enhances content richness and engagement, stimulating student enthusiasm and improving teaching quality. This paper examines the design of a multimedia teaching system incorporating virtual reality (VR) for engineering drawing. Key technologies such as graphics processing, 3D modeling, and dynamic databases are researched. Tools like Dreamweaver, Fireworks, SolidWorks, and 3DS MAX are used to create and integrate 3D models into web pages, forming a 3D model library. This library provides a flexible, content-rich, online learning environment that promotes self-study, enhances spatial imagination, and improves the quality of engineering drawing education. Establishing a virtual model library supports multimedia teaching and replaces traditional physical model rooms.
In crowdsourced testing, prioritizing numerous test reports is critical for improving developer review efficiency. Many researchers have proposed methods for prioritizing crowdsourced test reports for mobile applications, however, web crowdsourced test reports usually contain more detailed text and more complex screenshots, which makes prioritizing them especially necessary. This paper proposes a prioritization method named TCDiv, based on the data characteristics of web crowdsourcing test reports. First, using a pre-trained TextCNN model, the text of each report is segmented into two parts: reproduction steps and defect descriptions. Then, features are extracted from textual and image information respectively, and clustering technique is utilized to classify similar reports, and finally the clustering results are ranked and sampled. To validate the approach, experiments were conducted on 717 web crowdsourced test reports. The results show that TCDiv can detect more different defects in a limited time, thus improving the review efficiency of the development team.
To improve the quality of photovoltaic (PV) data and power prediction accuracy, a PV power prediction method based on neighborhood rough set and Pearson correlation coefficient (NRS-PCC) feature selection and multi-scale convolutional neural networks and long short-term memory (CNN-LSTM) network is proposed. We first calculate the correlation between different PV features based on PCC and select strongly correlated features to cross-multiply to get the fusion features to enrich the data source. Then, dimensionality reduction of the fusion features by NRS. Finally, correlation analysis based on PCC on the dimensionality reduction of the fusion features to screen out the effective features. Furthermore, a multi-scale CNN-LSTM is designed to predict PV power. The output vectors of different convolutional layers are first fused to extract multi-scale features, then the features of different scales are spliced as the input of the LSTM network, and finally, the LSTM network performs regression prediction. The effectiveness of the proposed method is verified on a real PV power dataset.
Based on the contradiction between the increasing power demand and the low capacity of the traditional direct current (DC) transmission system, this study innovatively combines the swarm intelligence optimization algorithm with power transmission and storage, and establishes a flexible DC transmission system model based on the swarm intelligence optimization algorithm. The study also analyzes the difference between the optimized capacity configuration of flexible DC transmission system and the traditional transmission system under different factors. The research results show that the transmission mode has become an important factor affecting the capacity allocation of flexible DC transmission system. The transmission medium and transmission cable type also have an impact on the capacity allocation of flexible DC transmission system. Optimizing power transmission mode, changing transmission medium and whether to add the swarm intelligence optimization algorithm are of far-reaching significance for improving the capacity allocation of flexible DC transmission system.
Objective To provide theoretical support for the study of AD pathogenesis and therapeutic targets. Methods The AD data were downloaded from the GEO database for differential expression analysis to obtain DEGs, followed by enrichment analysis of GO and KEGG signalling pathways, construction of machine learning models to screen key genes, and construction of risk prediction models and prediction of transcription factors based on key genes. In addition, consistent clustering analysis was performed on AD samples. Results Seven key genes were finally screened in this study, and the risk prediction model constructed on the basis of these seven genes had an AUC of 0.877. Cluster analysis classified the AD samples into two subtypes, and there was also a significant difference in immune infiltration between the two subtypes. Conclusion This study provides new perspectives and potential therapeutic targets for exploring the potential mechanisms by which mitochondrial autophagy affects AD, as well as providing directions for individualised treatment of AD.
Internet of things computer network technology can not only speed up the information transmission, but also help to break the previous time and space restrictions and improve the actual work efficiency. However, in the practical application and development of this technology, the realization of data security technology and remote control technology is the key to improve its application effect and ensure data security. Due to the obvious openness of the internet of things computer network itself, common security risks emerge one after another. Therefore, in practical application, it is necessary to pay more attention to security management and maintain network security.
Keeping pace with the times, this paper analyzes and interprets the concept, characteristics of Metaverse, and its influence on new media communication. Based on this, under the topological characteristics of information communication, this paper discusses the transformation of new media communication paradigm in the metaverse era and models the prediction of new media information dissemination in metaverse. In this paper, the multi-agent modeling method is used to analyze and compare the evolution results of each agent attribute under different parameter values. At the same time, based on the ideal social media network, the improved integrated immune control strategy is applied to the dynamic model of epidemic (SEIR) with latent period, and the spreading and evolution process of rumors before and after immunization is simulated. This research has both theoretical value and practical application value, which can provide a useful reference for the follow-up research of new media communication.
The prevalence of stress-related disorders has increased significantly in recent years, necessitating scalable methods to identify affected individuals. This paper proposes a novel approach utilizing large language models (LLMs), with a focus on OpenAI's generative pre-trained transformer (GPT-3) embeddings and machine learning (ML) algorithms to classify social media posts as indicative or not of stress disorders. The aim is to create a preliminary screening tool leveraging online textual data. GPT-3 embeddings transformed posts into vector representations capturing semantic meaning and linguistic nuances. Various models, including support vector machines, random forests, XGBoost, KNN, and neural networks, were trained on a dataset of >10,000 labeled social media posts. The top model, a support vector machine, achieved 83% accuracy in classifying posts displaying signs of stress.
