Design And Realization Of Text Clustering Prototype System

This paper provides the design and realization of a text clustering prototype system. The system is designed to meet the demands of an actual project which belongs to National Natural Science Foundation of China (NSFC).The evaluation of project applications needs many experts to evaluate lots of applications which is a very heavy work. Using the text clustering system can reduce the work intension, improve efficiency and save time. This paper designs the systematic architecture of text clustering system. And under this systematic architecture, it has discussed the analysis, design and realization of each subsystem in detail. The major completed works are as follows:(1) Lucubrates the partitioning methods in clustering algorithms, implements classic k-means algorithm and k-medoids algorithm which are used to clustering the project applications.(2)Lots of thesaurus and words that does not have class features reduce the clustering precision, so the system provides management of thesaurus and no-feature words, and improves precision.(3)After clustering analysis, label clustering results creates class models. Then classify new texts using class models.(4) User operation subsystem which is under B/S mode is developed by adopting Java and JSP technology. This subsystem regards JSP as control technology which is convenient to use, and provides visual graph of the results.