Collaborative geospatial feature search

Abstract

The ever-increasing stream of Web and mobile applications addressing geospatial data creation has been producing a large number of user-contributed geospatial datasets. This work proposes a means to query such data using a collaborative Web-based approach. We employ crowdsourcing to the fullest in that used-generated point-cloud data will be mined by the crowd not only by providing feature names, but also by contributing computing resources. We employ browser-based collaborative search for deriving the extents of geospatial objects (Points of Interest) from point-cloud data such as Flickr image locations and tags. The data is aggregated by means of a hierarchical grid in connection with an exploratory and a refinement search phase. A performance study establishes the effectiveness of our approach with respect to the amount of data that needs to be retrieved from the sources and the quality of the derived spatial features.

Automatic gazetteer enrichment with user-geocoded data

“…Others have mined places with their geographical extent from web pages [3], or mined places and determined their geographical extent from references such as postal code [32], or through a specially made tool such as the Jeocrowd to search user-generated data sets [19]. Others have asked users to help generate place names by setting up a separate web platform and asking people to add place names directly [31], or asking users to photograph every kilometer of the earth's surface and mine the place name tags indirectly, in the Geograph project 3 .…”

“…Nor is the aim of this research to build a comprehensive gazetteer resource, as did [22] and [1]. Our aim is to extract and compare the place names with a gazetteer, as did [19]. We determine the novelty of each entry for the purpose of integration into a gazetteer, as did [17], although the Kessler team did not consider the spatial precision of each extracted geo-tag.…”

Advanced techniques and algorithms to collect, analyze and visualize spatiotemporal data from social media feeds

“…1.2.1 Hierarchical grid/cube data structure Since our dataset includes single social media posts that are assigned an exact timestamp (second or millisecond precision) and detailed geographic coordinates (accuracy at the 100m level or more), we need a way to aggregate this information in order to be able to make sense of it. Initially ( [51]) we focused on the spatial dimension alone, overlaying the surface of the earth with a rectangular grid (mesh) of predefined cell size. By applying the grid on the dataset, individual social media posts get "bundled" into the grid cell that includes their geographic location, and the grid cell accumulates metadata for all the bundled posts.…”

“…Our work focuses on providing not only the location and the boundaries, but also a more comprehensive description of landmarks, places, neighborhoods or cities, by using social media feeds to create a heatmap of popularity within the extent of each feature. This chapter's content was initially included in [50] which was presented in the 19th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems in 2011, and later on in [51] which was presented in the 20th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems in 2012.…”

“…In the following, we describe the data structure used to efficiently compute some of these summaries. This so-called nanocube is an improvement of the hierarchical grid introduced in [51] that also handles the temporal dimension of the data. It is essentially a hierarchical cube (Figure 5.10) that can aggregate information in both space and time, while holding information about other dimensions, such as tokens, entities, etc.…”