The Research And Implementation Of Military Intelligence Retrieval System Based On Semantic Analysis

Electronic information has been applied in'military fields in deeply development of defense modernization, however, it has not been resolved how to quickly and accurately filter out useful information from mass data. Traditional data retrieval systems, no matter search engines or retrieve in limited field, are based on keyword retrieving, where the algorithm remains come down to superficiality of language rather than semantics, and lead the users hardly locating useful information. Latent semantic analysis contrived in this background, the method project characteristic into semantics by matrix singular value, describe semantic content of text from the point view of semantic, reduce the high-dimensional and sparse in some degree, and at the same time solve the problem of synonyms and near-synonym.Latent semantic makes use of potential structure of vocabulary and text, which are mapped to a K-dimensional semantic space, such a method can not only maintain most of original information, but overcame the phenomenon of multi-word, synonyms and word-dependent in the method of traditional vector space model. The efficiency of semantic space model is dependent to K, if K is too small it might lose some useful information, and if it is too large operation quantity will increase. In the practical applications, the value of K is defined by repeated experiments.The present work is to deeply study the latent semantic technology, and to extract original feature by the method, thereby to enhance the accuracy and recall rate of retrieval results; On the other hand, to improve the dependence of latent semantic on K and obtain optimization of sort of retrieval results by integrating sort results of latent semantic and vector space in the information retrieval stage.Ultimately, by the support of above research results, the issue develop retrieval system of military information, and successfully applied to automation of management in department. Experiments show that the system has a high practical value, and to some extent improve the accuracy and recall rate in the retrieval.