The Research And Implementation Of Text Similarity System Based On Power Spectrum Analysis

With the development of information technology in recent years, there are many kinds ofinformation around us. It’s very difficult for the clients to search the information they want.It’s very difficult to meet the clients’ demand with the original artificial management. For theexsisting Chinese text classification systems, although they can finish the classificationfunction, the clients are not completely satisfied with the results. Because of the particularityof the Chinese, it’s very easy to appear the wrong results thanks to the ambiguity in Chineseand the order of Chinese word.The paper focused on the analysises above and combined with the actual requirements,using different kinds of Chinese text classification as the background, put forward the ideathat we can use the method of power spectrum analysis based on multi-characteristics todistinguish the similarity of the Chinese text and complete text intelligent classification. Thespecific studies: research and implementation of Chinese text segmentation algorithm;research and implementation of text feature calculation and extraction and and building fieldthesaurus; the research and implementation of power spectrum matching algorithm andbuilding spectral libraries; the reseach of text similarity system and complete text intelligentclassification.The research and implementation of text similarity system based on power spectrumanalysis use the method of power spectrum analysis based of text multi-characteristics todistinguish the similarity of the Chinese text and finish text classification. The method is thatwe simulate the feature of brain signals in the writing process and combine the characteristicsof the Chinese word, and then set pulse signal function as text model, then get the powerspectrum of the text with the method of power spectrum analysis and build spectral librarybased on different areas. At last, we use power spectrum matching algorithm to distinguishtext similarity and finish text classification. The system uses the B/S (brower/server) structuremodel and SSH open sourse technology framework. The system uses the technology ofinvoking Matlab function from Java to get the coordinates of power spectrum and uses theJfreechart plugin to draw power spectrum.In conclusion, Chinese text similarity system finished the function of text intelligentclassification efficiently and accurately. The system could shorten the time that clients spendsearching for information and adapt to the social development.