Abstract: Event extraction is an important research direction in the area of information extraction. In response to the phenomenon that a large number of arguments arc missing in ACE event extraction caused by focusing on the current single sentence, we proposed a theory of filling the missing event arguments based on cross-event inference and achieved the prototype system. The system is divided into two parts that arc identification and classification of missing roles. Identification part is applied to decide whether a missing event argument can be filled while classification part is applied to decide which argument in other event mention can be used to fill the missing event argument. We annotated ACE 2005 corpus to reveal the filling relationships. I'he experimental results show that the F measure reaches 72. 97 and 74.68 respectively.

Key words: Filling of missing event argument, Cross-event inference, Event extraction