Research On The Method Of Power Spectral Analysis Based On Multi-features Of Text

With the rapid development of network technology and the ability of computersimulating the thought of human, the quantity of information is unceasing expansion.Artificial management information has been far from enough to meet the growing needs ofthe common. How to use the computer in the seemingly messy data quickly find potential,valuable information is current needs. In recent years, text similarity computing has gotconsiderable development, which has been widely applied to information retrieval,information filtering, machine translation, classified information and other fields, but mostresearch is studied on concrete application. In other words, one algorithm exist a poorapplicability in another field, which demands to study the new algorithm to meet the newapplications.In addition, systematic to represent text and the study of calculation methods are alsoexist many defects, hindering the development of text intelligent processing. Text calculationis one of the main theories about text intelligent processing. Furthermore, the mathematicexpression of text and its calculation is the basic method aiming at intelligent processing oftext. The paper by extracting text multiple values, build two dimensional feature set orientedto text and express text systematically. Two dimensional feature set is based on the set theory,which is vital significance in terms of the storage and application of text multi-values and thetext normalization processing. Through studying the composition, operational method and itsproperties for the two-dimensional features set, forming the calculation system oftwo-dimensional feature set based on text, which laid the foundation for the application of theword multi-values in Chinese.The paper has studied various classical algorithms of feature extraction and the textsimilarity calculation model. Through extracting multi-values from different perspectives,build economic domain subject thesaurus which is seen as core to research text energydistribution. In addition, the paper is enlightened by the idea that the human imaginationmovement will cause the change of brain waves. The pulse signals of brain caused by thewriting are closely linked with writing process. The paper builds the word pulse signalfunction by simulating the variation characteristics of the brain wave with the author writing processing and the contribution of some word values to text. And we can gain the text pulsesignal function by overlapping words pulse signal function. In order to solve the highdimensional of text, we convert the text pulse signal function to frequency domain, getting thepower spectrum graph of each text. Thus we propose the text similarity calculation modelbased on power spectrum estimation method. On the one hand, using the power spectrumgraph to represent text semantic and grammatical structures can get more, deeper semanticinformation features of text, increasing the accuracy of text representation, reducing the lossof semantic information. On the other hand, we are by the power spectrum analysis of the textto research the writing trend and internal law, to explore a new method of text analysis, toenhance precision and comprehensive of calculation.Finally, the paper by the formation of power spectrum library finished the judging of textsimilarity. The final experiment results showed that the power spectral matching algorithm isproposed in this paper can not only get rid of ambiguity between the language and charactersand the change of word order leading to error analysis results, but also finish the textsimilarity between the long text and long text, between the short text and short text, betweenthe long text and short text, which increases the application scope and effectiveness of textprocessing. At the same time, it also verified that by power spectrum estimation method torepresent text is feasible.