Causal Relation Recognition Between Sentence-Based Events

Identifying causal relations between events in news reports has always been an important research issue, as recognizing causal relations can help us get a clearer view of the event evolution, and also help making predictions and decisions. This paper proposes a method to recognize causal relations in news reports, which is combined of event extraction using trigger list, causal relation determination based on specific judgment rules, and finally a 2-dimensional SVM classification, where the features used for classification are extracted based on the context. Experiments which mainly focus on the new labor law related news reports have shown that this method works fine on the issue, and get a precision of 93% on an open corpus for test.