Abstract
Unstructured geo-text annotations volunteered by users of web map services enrich the basic geographic data. However, irrelevant geo-texts can be added to the web map, and these geo-texts reduce utility to users. Therefore, this study proposes a method to detect uncorrelated geo-text annotations based on Voronoi k-order neighborhood partition and auto-correlation statistical models. On the basis of the geo-text classification and semantic vector transformation, a quantitative description method for spatial autocorrelation was established by the Voronoi weighting method of inverse vicinity distance. The Voronoi k-order neighborhood self-growth strategy was used to detect the minimum convergence neighborhood for spatial autocorrelation. The Pearson method was used to calculate the correlation degree of the geo-text in the convergence region and then deduce the type of geo-text to be filtered. Experimental results showed that for given geo-text types in the study region, the proposed method effectively calculated the correlation between new geo-texts and the convergence region, providing an effective suggestion for preventing uncorrelated geo-text from uploading to the web map environment.
Highlights
Geo-text annotations are a type of map annotation that can describe map elements using unstructured texts, unlike traditional annotations consisting of structured words or numbers
POI geo-text [7] referring to users’ feedback on online goods, places, and services, such as the restaurant comments submitted by customers via Baidu or Google Maps, personal geo-text [8,9] referring to comments or descriptions that are attached to specific geo-positions and are shared by the volunteers to express their experiences, such as the personal tags provided by Google Earth [10] and the geo-text from Twitter presented in OpenStreetMap [11]
In order to prevent the addition of irrelevant geo-texts, semantic information in geo-text annotations and spatial autocorrelations within neighborhoods must be investigated according to the following principles: (1) The geo-text semantics are quantified by a text auto-classification method
Summary
Geo-text annotations are a type of map annotation that can describe map elements using unstructured texts, unlike traditional annotations consisting of structured words or numbers. Many geo-text annotation services based on crowdsourcing models have recently been provided by web map platforms [1]. In order to prevent the addition of irrelevant geo-texts, semantic information in geo-text annotations and spatial autocorrelations within neighborhoods must be investigated according to the following principles:. Voronoi diagrams assign point features or feature centers and establish boundaries between the separated objects, defining geometric adjacency between separated features based on boundary sharing [31] This method can automatically adjust for the spatial differences among irregularly distributed data points and uneven densities and is suitable for establishing a neighborhood for newly submitted geo-text annotations. The data quality in web maps has been reduced by the uploading of uncorrelated geo-text annotations, and filtering only illegal geo-texts is insufficient to correct the problem.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
