Automatic detection of discourse indicating emerging risk
Detecting emerging risk is a major concern in the "risk society" we live in. Risk can be detected among other sources from discourse describing events. Automatic language processing tools can help monitoring large amounts of electronic text, and recent advances in syntactic and semantic processing allow fine-grained analyses that produce normalized event descriptions, which can be used in risk detection. We have implemented normalized event extraction in the Xerox Incremental Parser.
We propose filtering all the normalized event descriptions in order to get events that indicate emerging risk based on two theories of detecting weak signals of emerging risk: one based on scenario models, implemented in the tool EventSpotter, and the other on detecting events that show characteristic features of weak signals.
In this article we describe the three modules (normalized event extraction and two ways of filtering) and propose them for industrial application as well as for social scientists involved in the analysis of discourse on risk.
Critical Approaches to Discourse Analysis across Disciplines http://cadaad.net/journal Vol 4 (2): 171 – 179
2009-025.pdf (351.89 kB)