Semantic Clustering for Large-Scale Documents.doc
Ming Liu · Computer and Information Science · 2010
Along with explosion of information, how to cluster large-scale documents has become more and more important. This paper proposes a novel document clustering algorithm (CLCL) to solve this problem. This algorithm first constructs lexical chains from feature space to reflect different topics which input documents contain, and documents also can be separated into clusters by these lexical chains. However, this separation is too rough. So, idea of self organizing mapping is used to optimize cluster partition. For agglomerating documents with semantic similarities into one cluster, influences from similar features are also considered. Experiments demonstrate that because effects of semantic similarities between different documents are considered, CLCL has better performance than traditional document clustering algorithms.