An approach to automatic interpretation of spatial descriptions based on large language models

G.A. Galimov; O.A. Nevzorova

doi:10.18127/j20729472-202601-05

An approach to automatic interpretation of spatial descriptions based on large language models

G.A. Galimov¹, O.A. Nevzorova²

research article UDC 004.8 Pages: 25-28 doi.org/10.18127/j20729472-202601-05

^1,2 Kazan Federal University (Kazan, Russia)

Keywords: Spatial reasoning large language models spatial relations spatial grounding spatial entity extraction

Abstract

Statement of the problem. Automatic interpretation of spatial descriptions is hindered by the ambiguity of spatial expressions and the lack of a formalized multi-step reasoning structure in LLM-based approaches.

Target. The aim of this study is to develop a formalized model of hierarchical spatial reasoning based on spatial operator composition.

Results. A model of sequential application of formally defined localization operators is proposed, enabling query interpretation to be represented as a computational process with an explicit structure. A classification of typical spatial grounding errors is introduced, and the advantages of explicit reasoning representation over implicit agent-based strategies are demonstrated.

Practical significance. The approach is applicable to geographic information systems and intelligent navigation systems.

Received: 24.02.2026
Approved after review: 26.02.2026
Accepted: 10.03.2026

How to cite

Галимов Г.А., Невзорова О.А. Подход к автоматической интерпретации пространственных описаний на основе больших языковых моделей // Системы высокой доступности. 2026. Т. 22. № 1. С. 25−28. DOI: https://doi.org/10.18127/j20729472-202601-05

References

Palanichamy N., Maheswar R., Trojovský P. GeoNLU: Bridging the gap between natural language and spatial data infrastructures.
Alexandria Engineering Journal. 2024. V. 87. P. 126–147. DOI: 10.1016/j.aej.2023.12.027
Kuhn W. Geospatial semantics: Why, of what, and how? Journal on Data Semantics III. Lecture Notes in Computer Science. V. 3534. Berlin; Heidelberg: Springer. 2005. P. 1–24. DOI: 10.1007/11496168_1
Annepaka Y., Pakray P. Large language models: a survey of their development, capabilities, and applications. Knowledge and Information Systems. 2025. V. 67. P. 2967–3022. DOI: 10.1007/s10115-024-02310-4
Russell S., Norvig P. Artificial Intelligence: A Modern Approach. 4th ed. Munich: Pearson. 2021. 1168 p.
Zheng S., Fang M., Chen L. SpatialWebAgent: Leveraging Large Language Models for Automated Spatial Information Extraction and Map Grounding. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (ACL). 2025. P. 252–266. URL: https://aclanthology.org/2025.acl-demo.25.pdf

[1] Palanichamy N., Maheswar R., Trojovský P. GeoNLU: Bridging the gap between natural language and spatial data infrastructures.
Alexandria Engineering Journal. 2024. V. 87. P. 126–147. DOI: 10.1016/j.aej.2023.12.027

[2] Kuhn W. Geospatial semantics: Why, of what, and how? Journal on Data Semantics III. Lecture Notes in Computer Science. V. 3534. Berlin; Heidelberg: Springer. 2005. P. 1–24. DOI: 10.1007/11496168_1

[3] Annepaka Y., Pakray P. Large language models: a survey of their development, capabilities, and applications. Knowledge and Information Systems. 2025. V. 67. P. 2967–3022. DOI: 10.1007/s10115-024-02310-4

[4] Russell S., Norvig P. Artificial Intelligence: A Modern Approach. 4th ed. Munich: Pearson. 2021. 1168 p.

[5] Zheng S., Fang M., Chen L. SpatialWebAgent: Leveraging Large Language Models for Automated Spatial Information Extraction and Map Grounding. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (ACL). 2025. P. 252–266. URL: https://aclanthology.org/2025.acl-demo.25.pdf

Abstract

How to cite

Related articles