Soft data analysis within a decision support system

This paper presents a work achieved within a decision support system. The aim is to support an operator in his/her task of analyzing soft data to monitor and anticipate a geopolitical crisis. The incoming data is filtered and analyzed so as to retain only the relevant events. Therefore, the system extracts the relevant events from the incoming soft data and produces synthetic descriptions of these events. The approach is divided into three steps. First the domain of interest is modeled with an ontology. It describes the events that are relevant to monitor, as well as geographical and political concepts and issues. Then, the system extracts the relevant information items from texts, based on linguistic patterns. Finally, the extracted items are fused to deduce synthetic events. The fusion process is based on conceptual graphs. The approach is illustrated using a collection of news and newspaper articles that deal with Ivory Coast.