A POI selection method based on GCN considering annotation conflicts during map scale transformation

International Journal of Geographical Information Science, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 31

Published: Feb. 27, 2025

Language: Английский

---

Multi‐Row Labeling With Semantic Analysis: A Case Study on Chinese POIs

Transactions in GIS, Journal Year: 2025, Volume and Issue: 29(2)

Published: March 24, 2025

ABSTRACT Labels are widely used in maps to convey verbal information for symbols and play a crucial role aiding users' navigation understanding of spatial context. Traditional labeling approaches mainly focus on ensuring label readability by placing them overlap‐free while maintaining visual coherence. However, these methods often fall short Point Interest (POI) labeling, particularly the context Chinese labels, due growing demand detailed informative with long or descriptive POI names. To address this challenge, it is essential shorten names split labels into multiple rows achieve more effective layout. In work, we present multi‐row algorithm that introduces new quality constraints incorporates linguistic semantic analysis preprocessing, segmentation, placement. We also demonstrate its application real‐world mapping platform, Meituan Map, which serves over 50 million monthly users. inform design algorithm, interviewed six domain experts conducted statistical based dataset containing 160,297 POIs. This confirms necessity highlights practical challenges considerations. Our results indicate preprocessing phase our can reduce total character count 41.15% word 43.24%. Comparative experiments show approach achieves superior placement terms clarity. improvement may come at cost increased computational time. A user study involving 318 participants demonstrated processing result user‐preferred layout comparable effectiveness tasks, although increase response

Language: Английский

---

User portrait analysis on Chinese diabetes online health platforms: an information processing perspective

Online Information Review, Journal Year: 2025, Volume and Issue: unknown

Published: April 14, 2025

Purpose The objective of this research is to investigate the characteristics information interaction among users largest online health platform for diabetes in China from an processing viewpoint, determine stages and reveal variations requirements behavioral patterns across different user groups at various levels, ultimately creating a segmentation labeling system enhance portrait. Design/methodology/approach This study adopts deep learning BILSTM-CNN classification model identify characteristics, then classify into three groups. LDA topic employed analyze needs these Findings utilizes combined model, showcasing enhanced effectiveness classifying degree comments. Our also increases accuracy stability compared conventional models, achieving F1 score 95.0% (F1 Score: CNN 92%, LSTM 94%, BILSTM 94%). Based on results, were grouped showed differences needs, behavior natural attributes. Originality/value Taking as entry point, deeply mined data China’s platform, identifying present comments categorizing reflecting varying depths processing. multi-dimensional analysis, we innovatively constructed refined system, finally depicted complete not only enriches theoretical framework cognitive portrait but contributes personalized recommendations platforms diabetes. Peer review peer history article available at: https://publons.com/publon/10.1108/OIR-11-2024-0728 .

Language: Английский

---

A continuous merging method for natural surface elements based on morphing

Geocarto International, Journal Year: 2025, Volume and Issue: 40(1)

Published: April 21, 2025

Language: Английский

---

Building pattern recognition by using an edge-attention multi-head graph convolutional network

International Journal of Geographical Information Science, Journal Year: 2024, Volume and Issue: 39(4), P. 732 - 757

Published: Nov. 18, 2024

Effective building pattern recognition, a complex task that requires the simultaneous consideration of individual features and spatial relations, is essential for successfully generalizing maps. However, existing deep learning approaches must still be adequately comprehensive in jointly quantifying relationships buildings, suggesting further improvement quantitative representation spaces. This study presents novel edge-attention multi-head graph convolutional network (GCN) concurrently considers modeling enhancing recognition. The proposed method captures including proximity arrangement similarity, by using relationship descriptors attention mechanisms to generate relevance coefficients. These coefficients are then integrated into weighted GCN participate expression features, facilitating analysis thus, improving recognition performance. Our experimental confirms method's superior capability recognizing features. also demonstrates strong generalization across different scales areas, underscoring its efficacy potential geospatial analyses.

Language: Английский

---

A Machine Vision Approach to Real-Time Drone Detection Using Gray Level Co-occurrence Matrix and Interactive GUI

2022 International Telecommunications Conference (ITC-Egypt), Journal Year: 2024, Volume and Issue: unknown, P. 554 - 560

Published: July 22, 2024

Language: Английский

---

Investigating the response of urban surface thermal environment to urban morphology changes in Nanjing

Journal of Environmental Management, Journal Year: 2024, Volume and Issue: 370, P. 122875 - 122875

Published: Oct. 16, 2024

Language: Английский

---

Triangular Mesh and Neural Network for Object Search Based Cluster Centre Descriptor

SN Computer Science, Journal Year: 2024, Volume and Issue: 5(8)

Published: Dec. 14, 2024

Language: Английский

---