Research On Journal Decision-making Reference Based On Text Mining

As the country attaches increasing importance to innovation,the accumulative amount of various periodicals also increases continuously.Analyzing and excavating the effective knowledge of periodical industry is of great significance for self-assessment,author evaluation and article quality evaluation.In order to provide a reference for decision-making of periodicals,this paper proposes a text similarity measure method based on classification lexicon and a topic text extraction method based on keywords.Taking the above two methods as the main research methods,this paper combines similarity measure methods of text,clustering,classification methods to complete periodicals decision reference research.Aiming at the limitations of high time complexity of existing text mining based on semantic knowledge rules analysis,a text similarity measure method based on thesaurus is proposed.Regarding “Modern Chinese Classification Dictionary” as the semantic knowledge,this paper improves the method of similarity measure of words based on existing similarity measure methods.Other similarity measure methods based on semantic knowledge are selected as comparative experimental methods.The results of the similarity measure method are verified by clustering and classification experiments,which proves the rationality of this method.The method uses the thesaurus as the knowledge base.Compared with Ci Lin and CNKI,the thesaurus contains more words.Thus,the matching success rate of word coding is higher and the similarity of the text is less affected.In the process of calculation,the knowledge base is accessed only in the phase of word coding match,which improves the time efficiency of the method based on semantic knowledge base.A new method of word similarity calculation is proposed,which is superior to other methods based on semantic knowledge base.In view of the current research status of document analysis by researchers who generally use statistical methods to analyze documents,text similarity measure based on semantic knowledge base and periodical text topic extraction method based on keywords are used for knowledge discovery of the journals.This paper considers the similarity measure method based on thesaurus as the theoretical basis and papers from 2007 to 2016 in the kind of creation as the research object.Cluster the documents automatically for times through AP clustering.According to the clustering level,the key words are clustered,topic words extracted and transformed.The final theme extraction results are presented in the form of theme tree.Building a theme tree for the theme extraction results,which visually displays the theme extraction results,and clarifies the topic frame of the research object.Taking the literature key words as the subject of literature,to a certain extent,avoids the limitation that the words extracted in the literature can not express the subject of literature.Using the word similarity measure based on the semantic dictionary,the errors caused by the synonyms are reduced to some extent.On the basis of the above two theoretical methods,this paper makes a comparatively deep analysis of the periodicals of the target journal from the aspects of subject analysis of journal articles,citation-reference relevance,organization / fund-theme,aiming to provide certain reference for journal decision-making.In the journals decision-making reference study,the research topics of the target journal periodicals in 2011 ~ 2016 are extracted and summarized,and the main research topics in different annual journals and the research topic change trends in annual journals can be clearly grasped.Based on the similarity measure of words,the citation-reference correlation is estimated,and the relevance of the journal reference is generally understood.In the meantime,the citation rate of the journals is increased while the citation itself is quoted as the quality of other essays.The conclusions of the research on the citation frequency and the main research topics of different agencies / fund issue papers are provided,and some references for decision-making are given to the hiring of journal articles and peer review.