Intelligent monitoring of industry-education practical training requires reliable recognition of teaching events involving people, equipment, and workstations. However, visually similar devices, occlusion, idle equipment, and incidental proximity often create ambiguous evidence. This study proposed an equipment-relation guided framework for teaching behavior detection. The model detected persons, equipment, and workstations and formed relation units encoding spatial affiliation and instructional relevance. An equipment participation gate integrated operation responses, relation priors, workstation consistency, and background-confusion cues, while a short-term temporal module improved robustness to occlusion and transient motion. The equipment-relation guided framework recognized teacher demonstration, student operation, teacher guidance, student observation, and risky operation. On the two datasets developed in this study, it improved F1-score by 5.99 and 5.96 percentage points and mean average precision@0.5:0.95 by 6.73 and 6.78 percentage points over the baseline. Ablation, significance, complexity, and visualization analyses confirmed its effectiveness.
Deep learning detectors for online misinformation report near-perfect in-domain accuracy, yet deployment faces distribution shift. This study asks whether such scores reflect a transferable ability to separate true from false content, auditing across three public corpora needing no social-media interface: LIAR statements, McIntire news, and ISOT news. A controlled ladder (term-frequency linear classifiers, a bidirectional LSTM, and fine-tuned DistilBERT and RoBERTa) is evaluated in-domain and across every corpus pair, with three-seed statistics, calibration, and interpretability. Both news corpora are nearly saturated in-domain (RoBERTa 0.999 ROC-area under the curve), while the short-statement corpus is hard for every model (0.657). Transfer is governed by genre: news-to-news transfer is strong (0.943) and statement-to-news is recovered by pretraining (0.935), but transfer into the statement corpus stays at chance (at most 0.527). Vocabulary overlap tracks these outcomes, and linear detectors grow overconfident under shift. Therefore, in-domain success gives weak assurance against distribution shift.
The growing linked and multidimensional aspects of data on the web create challenges for analysts who need to balance their exploration and exploitation of such data sources. However, this balance is key to unlocking more value out of the data to meet organizational objectives. To address this challenge, the authors propose a framework designed to enable a more balanced web of data exploration and exploitation through a Personal Knowledge Graph Cube ontology. They evaluate the proposed approach through an online prototype developed to enable analysts to construct personalized knowledge graph cubes and reuse the cubes in external analytical tools. The prototype was used by 80 analysts who created 520 cubes from a repository of 482,000 RDF triples. Initial results of a mixed-methods evaluation revealed stable system performance and data manipulation times. Interviews and usability feedback from 31 analysts showed promising System Usability Scale score of 82.18, with qualitative insights of the system’s ability to support more personalized exploration-to-exploitation workflows on the web of data.
In complex traffic environments, conventional vehicle detection methods often show limited precision and robustness when facing distant, small-scale, and occluded vehicles. To address these issues, this research proposes a Multiscale, Multi knowledge, Attention enhanced You Only Look Once (YOLO) model. It is a knowledge-driven multiscale vehicle detection framework for intelligent transportation systems, built on the YOLO version 8 nano model. The framework introduces three modules. The first is a cross-stage partial mixed aggregation network module. This module enhances backbone representation through dynamic multiscale aggregation. The second is a multiscale downsampling module that combines dilated convolution and parallel pooling to preserve vehicle cues during downsampling. The third module is an aspect-ratio perception with cross-attention module that injects aspect-ratio-aware knowledge with cross-attention to adapt to vehicle shape variation and suppress background interference. Experiments on the Vehicle dataset on Kaggle.com and the Berkeley DeepDrive 100K dataset showed that the Multiscale, Multi knowledge, Attention enhanced-YOLO model outperformed the lightweight YOLO baseline models. On the Berkeley DeepDrive 100,000 dataset, it achieved a 48.17% mean average precision @0.5 and 26.70% mean average precision @0.5:0.95, improving YOLO version 8 nano by 3.34% and 2.19%, respectively. Visualization results further confirmed its effectiveness under occlusion and adverse-weather conditions while maintaining real-time efficiency.
Blockchain gaming has evolved from recreational play into a profit-oriented digital activity. However, the factors driving sustained player engagement in gamified finance remain insufficiently understood. This study proposes an integrated model combining the Theory of Planned Behavior and the Value-Intention Framework. Data were collected from 203 valid respondents recruited through Taiwanese online blockchain gaming communities. A two-stage analysis was employed, using structural equation modeling and artificial neural networks. The structural equation modeling results show that attitude, subjective norms, flow experience, perceived ease of use, and reward mechanisms significantly influence continuance intention, whereas perceived behavioral control and perceived risk do not show significant linear effects. Artificial neural network sensitivity analysis further identifies attitude as the strongest predictor and highlights the relative importance of perceived behavioral control, suggesting potential nonlinear mechanisms. The findings provide theoretical and practical insights into player continuance in gamified finance.
