Research On Chinese Automatic Summarization Based On Clustering Algorithm

With the development of search engine technology, people can use various search engines to find the information they need. People enter some keywords and the search engine will return the related web pages. Unfortunately, search engines cannot return concise answers to the users. So automatic summarization technology becomes more and more important, and it has an extensive application prospect.As to solve the redundancy problem in Chinese automatic summarization, we combine latent semantic analysis (LSA), HowNet and senence clustering algorithm to generate Chinese automatic summarization. We use LSA and HowNet concept extraction to compute the similarity of sentences, which increase the accuracy of sentence similarity computing.On the other hand, we do not use a single hierarchical clustering algorithm or partition clustering algorithm, instead we use a mixed algorithm instead, which mix both hierarchical clustering algorithm and partition clustering algorithm. The experiment shows that the hyper clustering algorithm has a better result.We design and develop a prototype to verify the methods represented in this dissertation, and design two experiments using this system. Experimet one compares the precision and recall rate of the two sentence similarity computing methods. Experinet two compares the clustering accuracy of three clustering algorithms. In the end, we analysis the experiment results in detail.