Design And Implementation Of The Knowledge Management System Based On Text Classification

With the continuous development and mature of network technology, more and more organizations and institutions join the Internet, post and share information on the Internet. As a result, the way of people management information is convert from books to the Internet, but the amount of information on the network is rapid growth every day,information available on the Internet is manifolds of numerous and complicated and varied forms. This brings diversity for information acquisition. Demand for these users efficient organization and management of information and converted to their knowledge,increased the difficulty of knowledge management.Organize and manage large amounts of information that collected by the user has become a research hot spot. Text classification is an important form of text data analysis,it builds the important text data type model, classification is used to predict the class label. Text classification can be effective to collect information form different categories,and convenient for the user to obtain what they need in the knowledge management system of information.In this thesis, based on the study of text classification, considering the impact of the classification process at every stage of the configuration of different parameters on the classification results, build a kind of adjustable and optimal classifier. In view of the different of existing knowledge management system design and their respective applicable environment, apply the text classification to the knowledge management system, design and implement suitable for the individual application of knowledge management system, effectively to organize and manage the user’s knowledge. The whole system is divided into knowledge acquisition subsystem, knowledge processing subsystem, knowledge display subsystem, and complete their respective responsibilities.Innovation point of this thesis related work is mainly in the following two aspects:(1)Firstly we obtain clusters by text clustering in the process of using user’s sets of text knowledge to construct a text classification datasets. Then, we need to extract keywords for each cluster labeling. In this thesis, candidate words are composed by title words after title participle and high value of TFIDF words in the clusters after clustering.Then, consider the location information of the word and give it weight, proposed a method to calculate the representative of the candidate word and select the highestrepresentative value as the cluster label. Thus construct a complete classification datasets.(2)The algorithms that used to clustering and used in the classifier building process are run paralleling by use Map Reduce.