In this paper, we propose a novel model for unsupervised Chinese event extraction. We use a multi-information fusion technique to combine two kinds of information for knowledge representation of event instances: language features and structure information. Then, we perform our proposed XLS-means Clustering Algorithm to group the candidate event instances into a "natural" number of clusters, which can fully take into account the similarity of both their language and structure information. The experimental results on ACE2005 Chinese corpus show that our model can achieve better performance than other unsupervised methods.
展开▼
