Geographic information extraction from user contributed data Full text

Savvas Georgiou
School of Electrical and Computer Engineering, NTUA
Diploma Thesis
Abstract. Geospatial data has become an important resource in todayʼs Web applications not only as content but also as metadata. Despite its undisputed usefulness, issues need to be addressed with respect to the availability, the accuracy and the cost of the data. Geospatial data has typically been generated and, thus, curated by professionals, e.g., surveying firms and large map data providers. The advent of Web2.0 created several creative-commons initiatives addressing geospatial dataset creation. In addition, countless (mobile) applications have been producing large amounts of point-ofinterest (POI) datasets. In this work, we show how to exploit such user-contributed data by deriving integrated geospatial datasets from user-contributed data. We will give methods for querying and integrating user-contributed data based on spatial as well as lexical clustering. An experimental evaluation will establish the applicability and quality of the approach in terms of obtained datasets.