Interpreting natural language descriptions of the topological relations of enclaves

• Published in: Journal of geographical systems Vol. 27; no. 2; pp. 301 - 335
• Main Authors: Wang, Xiaonan, Zhang, Xiuyuan
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.04.2025; Springer Nature B.V
• Subjects: Computer Appl. in Social and Behavioral Sciences; Configurations; Descriptions; Econometrics; Economics; Economics and Finance; Geographical Information Systems/Cartography; Landscape/Regional and Urban Planning; Language; Natural language; Original Article; Regional/Spatial Science; Topology; Urban Economics; Y91; Incompatible expressions; Enclaves; Natural language descriptions; Topological relations
• ISSN: 1435-5930; 1435-5949
• DOI: 10.1007/s10109-025-00461-8

An enclave is a part of country that is surrounded by another country, or any small, distinct area or group enclosed or isolated within a larger one. Enclaves' spatial configurations often influence local cultures, national politics, and even international relationship, but they are much more complicated to measure and express compared to non-enclaves, for example, natural language expressions used to describe enclaves' spatial relations can include the mutually exclusive terms, e.g., "inside" and "surrounded by". To better understand natural language expressions of enclaves, this research investigates whether preferences exist for various spatial relation terms and assess the factors that may influence the choice of different expressions used to describe enclaves' spatial configurations. Using human subject testing, we quantitatively evaluated 18 spatial configurations, four categories, and rotation and shape changes applied to enclave representations, and asked subjects to select from among five spatial relation choices and five predefined reasons for their choices. The survey results from 126 participants provided the basis to construct a random forest model for predicting enclave spatial relation choices, given different spatial configurations and contexts, which has an acceptable prediction accuracy of 77.04% and indicates Reasons that play a more important role (76% importance) in predicting the choice compared to classical spatial configuration measurements. It demonstrates the natural language descriptions that usually differ from the formal topological models. These results facilitate consistent interpretation of natural language used to describe enclaves, and support further studies on employing Reasons as an influential factor for natural language spatial queries.