Knowledge-based Highly-specialized Terrorist Event Extraction

In this paper we present a prototype of a system aimed at event extraction using linguistic patterns with semantic classes. The process is aided with an auxiliary tool for mapping verb statistics across messages. The sentence analyzer uses linguistic associations, based on VerbNet across the message and between messages' sentences to select semantic role fillers. We restrict ourselves to the coverage of one event type only – namely a kidnapping  and to two events template slots (semantic roles): a perpetrator and a person_target (a human target). We designed rules involving semantic role filling using previous works on coreference. We used the Sundance parser and AutoSlog extraction patterns generator. Then we applied the semantic role filler and event resolution tool SRL Master. Our approach yields high performance on the MUC-4 data set.