Design and implementation of Chinese words clustering based on atomic-concepts

Cluster analysis is an important technique in statistics; it is an effective way to find hidden knowledge behind huge amounts of data. Especially now, in one hand, the amount of all kinds of papers is too huge to read; in the other hand, the extensive use of search engine technology also provides a huge number of words to find all kinds of information, and thus how to find the information from these words using words cluster become a meaningful issue. The method of words cluster in this paper is based on a certain improvement to the previous methods, which is the idea of putting forward atomic concepts, by calculating the similarity between the source words and the atomic concepts and weighted to obtain the clustering space. Then we use the FCM cluster algorithm of Matlab to calculate and achieve relatively good results.