EUVIMEDEUVIMED entdecken
Crossref

An Iterative Framework for Extracting Nested Spatial Relation Patterns From Geoscience Texts

Deping Chu, Siwen Chen, Bo Wan, Fang Fang, Shunping Zhou, Wei Luo, Lin Yang

Transactions in GIS · 2026-11 · 30 (7)

DOI: 10.1111/tgis.70457 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Transactions in GIS, online erschienen am 06.11.2026. Verfasst von Deping Chu und 6 weiteren Autor:innen. Die Arbeit nennt 45 Literaturangaben. Gefördert von: National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Spatial relation extraction (SRE) aimed to identify spatial relations such as direction and topology between entities described in text. However, existing methods overlook the recognition of nested spatial relation patterns, in which one spatial relation may contain another complete spatial relation unit. To address this limitation, we introduce Geo‐RE, a novel framework designed to iteratively extract nested spatial relation patterns by effectively integrating both semantic and structural information from geoscience terms. Our approach leverages the pre‐trained language model Geo‐BERT, which captures the rich contextual semantic information specific to geoscience texts. Additionally, a convolutional neural network (CNN) is employed to extract radical features of Chinese words, which aids in identifying structural and semantic characteristics unique to geoscience terminology. During the iterative hierarchical extraction process, a candidate relation triple filtering mechanism is designed to eliminate semantically or physically invalid relations, thereby ensuring identification efficiency and accuracy. Finally, a hierarchical neural network is adopted to iteratively identify nested spatial relation patterns in a layer‐by‐layer fashion. Experimental results on both the Policy dataset and our proprietary GeoRED dataset demonstrate that Geo‐RE achieves an optimal F1‐score of 88.10% on GeoRED and performs comparably to state‐of‐the‐art models on the Policy dataset.

Förderung

  • National Natural Science Foundation of China

Zitieren

APA

Chu, D., Chen, S., Wan, B., Fang, F., Zhou, S., Luo, W., & Yang, L. (2026). An Iterative Framework for Extracting Nested Spatial Relation Patterns From Geoscience Texts. Transactions in GIS, 30(7). https://doi.org/10.1111/tgis.70457

Vancouver

Chu D, Chen S, Wan B, Fang F, Zhou S, Luo W, et al. An Iterative Framework for Extracting Nested Spatial Relation Patterns From Geoscience Texts. Transactions in GIS. 2026;30(7). doi:10.1111/tgis.70457

BibTeX
@article{Chu2026,
  author = {Chu, Deping and Chen, Siwen and Wan, Bo and Fang, Fang and Zhou, Shunping and Luo, Wei and Yang, Lin},
  title = {An Iterative Framework for Extracting Nested Spatial Relation Patterns From Geoscience Texts},
  journal = {Transactions in GIS},
  year = {2026},
  volume = {30},
  number = {7},
  doi = {10.1111/tgis.70457}
}
Kostenlos starten →