Geospatial Database Generation Agents

We propose to develop technology for geospatial database generation (GDBGEN) by combining six key components: 1) Use of intelligent crawling to identify and efficiently process documents on the World Wide Web relevant to a specific location or geographic feature 2) Spatio-temporal parsing using on new innovations to recognize and resolve location names and time expressions and an automated recognition engine built using conditional random fields 3) Text parsing using natural language processing to parse texts into appropriate lexical, syntactic, and semantic units to identify a wide range of descriptive features that may be expressed in tables or natural language sentences 4) A coordination component to seamlessly relate locations, times, and descriptive features to each other both within and across document boundaries into a collection of semantic “geospatial knowledge structures” 5) A reasoning component which provides a robust broad coverage framework for assigning confidence scores and selecting the best factoids from a set of potentially conflicting candidates 6) A data management layer which facilitates user interaction on multiple levels of granularity and enables discovery, visualization, and export of data using open standards.