From Text to Semantic Geodata Enrichment

Geographically referenced information are captured, managed, analyzed and displayed by the geographic information system GIS. This information will help to better understand and analyze a host of human activities in an attempt to reach a more sustainable future. Moreover, these information are sometimes hardly available due to the cost entailed by the acquisition process and the maintenance. Hence, efforts are devoted to bring the interesting information with less cost. In this context, the authors propose an approach dedicated to the enrichment of the semantic aspect of geographic database GDB. The approach emerges from the agents' interaction in conformity with the multi-agents framework. It deals with an Interface agent, Task agents and Geographic ones. These agents are to process a corpus of documents related to the geographic entity at hand. This processing consists of extracting the essence of information from the managed corpus to enrich the GDB. These agents follow a pipelined approach consisting of the following stages: text fragmentation, thematic annotation, delegation and text filtering. Besides these basic stages, a refinement is to be performed in case of user unsatisfaction and this by exploring the spatial relationships.