Research And Realization Of Automatic Classification Of Military Information

As the pivotal technique of military information (MI) intelligentized processing, The automatic categorization of Ml is the indispensable means to improve the command efficiency, which is studied in this paper including obtaining the MI training set, building the denotation model of MI, determining the MI attribute, choosing the categorization model of MI and evaluating the categorization performance. The main creative work of this paper are concluded as follows:1. Through analyzing the denotation model of general text and combining the character of Ml based on XML, the paper improves the original weight calculation method of VSM (vector space model) and presents the novel MM (Maximal Match) algorithm.2. Aiming at the problem of trending to the low-frequency words existing in original mutual information method, we present an improved attribute-choosing algorithm of MI.3. The system introduces the feedback module into the simple VSM categorization algorithm, and associates the vector distance method with the feedback method to improve the categorization precision greatly.4. After integrating the above algorithms and models, we design and realize the MICS (Military Information Categorization System), buildup the total construction and research the relationship between the modules, meanwhile get the testing and evaluation of the system performance.