This proposal addresses topic A07-124 ``Geospatial Database Generation Agents'' aiming at developing the technology to populate the geometry and attribution of geospatial databases with information mined from open source text, tables, and non-spatial databases on the Web. As our solution, we propose to build the GeoEngine, a ``vertical-search'' system for finding and ranking information about geospatial features from both the ``surface'' and the ``deep'' Web, by extracting textual properties and patterns that identify geospatial objects and their attributes, indexing their occurrences, and integrating such occurrences over numerous Web pages, in order to search for pertinent information. We will build an end-to-end geospatial database generation agent system for assisting users and analysts to gather geospatial information from the open Web.
Keywords: Geospatial Database, Web Search, Information Integration, Data Extraction