Towards the Ordering of Events from Multiple Textual Evidence Sources

Abstract

In any criminal investigation, two important problems have to be addressed: (1) integration of multiple data sources to build a concise picture of the events leading up to and/or during the execution of a crime, and (2) determining the order in which these events occurred. This paper focuses on the integration of multiple textual data sources, each providing a recollection of events observed by eyewitnesses. From these textual documents, using text mining and natural language processing techniques the authors identify events (across the document corpus) associated with the crime and infer temporal relationships between these events to create a (partial) ordering of the events. The authors evaluate their method on data collected through a mock eyewitness task.