Utilizing the Structure and Data Information for XML Document Clustering

This paper reports on the experiments and results of a clustering approach used in the INEX 2008 Document Mining Challenge. The clustering approach utilizes both the structure and the content information of the XML documents in the Wikipedia collection. The content of the XML documents is measured using the latent semantic kernel (LSK). A well-known problem with the construction of latent semantic kernel is the use of singular vector decomposition (SVD) method on a large feature space which is extremely expensive, in terms of computational as well as of memory requirements. Therefore a dimensional reduction method based on the common structural information of the XML documents is applied to reduce the dimension of the document space for building the latent semantic kernel. After the kernel is constructed, the proposed clustering approach uses the kernel to measure the similarity between each pair of document contents in the dataset. The proposed clustering approach has shown to be effective on the Wikipedia dataset.