Research And Implementation Of Topic-Based Event Fusion

With the rapid development of the Internet, the information about certain topics has been increasing explosively. Due to the redundancy, incompleteness, inconsistence and ambiguity of the element information, it becomes an urgent on how to acquire the detailed and complete information upon a certain topic. This paper focuses on the topic event fusion, and the main fruits are listed as follow:Firstly, this paper explores the fusion of multi-source and imperfect event information. In related to the absence and incompleteness of the event information, a statistics based element supplement approach is proposed. On the improvement of the current element similarity and clustering algorithm, the recall and accuracy on the element clustering are also raised. Then it proposes a meta-element group fusion strategy, which can effectively solve the fusion of exclusive, complementary and redundant information.Secondly, we deal with the topic event fusion, and propose a clustering algorithm on word co-occurrence element. By incorporating the topic based elements effectively, it proposes the abstract generation method based on the element, as well as the hierarchical clustering algorithm based on abstracts, and produces the structured and temporal formalized elements.The experimental results show that the event fusion method is useful to fuse the event mentions and organize the relative events. It can reduce the information redundancy sharply and then consummate the event information. By fusing the redundancy and complementarity of the multi-source information, it has successfully added the dimension of the target feature vector, and reduced the information definiteness, and improved the confidence level of the information. The topic fusion method is also useful to contact relative events, and organize the topics in hierarchy and structure form.