Propagation and Visualization of Uncertainty in NL-Based Spatial Analysis

Abstract

Natural language (NL) as the most natural expression way of human is an ideal approach in spatial analysis. A growing attention of GIS scientists and AI scientists has been paid to the uncertainty of NL-based spatial queries and spatial analysis which include uncertainty of natural language expression and processing as well as the one of GIS model and spatial data source. The uncertainty of natural language expression that origin from vagueness of natural language itself, limitation of spatial cognitive ability of speakers and incompleteness of natural language expression will probably be brought into natural language processing result. In spatial analysis calculation based on natural language, the calculation result uncertainty is from uncertainty of starting feature, uncertainty of operator and uncertainty of calculating factor. The confidence of spatial analysis result is a mixture of the uncertainty sources listed above. Furthermore the visualization of spatial analysis uncertainty as another hot issue in uncertainty research field helps domain experts and other users realize reliability of spatial analysis results plainly. In this paper focusing on the study of propagation and visualization of uncertainty in NL-based spatial analysis, firstly, we summarize the uncertainty source of NL-based spatial analysis especially in NL-based location determination systems, secondly, we set up an uncertainty propagation model of NL-based spatial analysis and take an example of variant confidence distribution in variant part of linear and regional features in spatial analysis based on natural language, thirdly, we design a plain uncertainty visualization in GIS GUI, finally, an experimental prototype was developed to verify models, analysis result and visualization design.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 3 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.