How text-mining could improve surveillance systems?

Source Agritrop Cirad (agritrop.cirad.fr) International audience The ability to identify emerging and re-emerging diseases is challenging for the health domain. In this context, event-based surveillance (EBS) gathers information from heterogenous data sources, including online news articles. EBS systems integrate text-mining methods to deal with huge amounts of textual data. This talk focuses on the use text-mining and multidisciplinary approaches in order to mine news data dealing with the health domain. These data science approaches are integrated in an EBS system called PADI-web (Platform for Automated extraction of Disease Information from the web). PADI-web dedicated to animal health surveillance tackles disease-based and symptom-based surveillance. To address these issues different text-mining methods associated with labeled textual datasets are integrated in the main steps of EBS systems: data acquisition, information retrieval (i.e. identification of relevant texts), epidemiological information extraction, information to communicate to end-users. These methods are also adapted in other domain like Food security by mining heterogenous data.