The Research Of Co-word Analysis Based On Mixed Weighted

With the increasing number of journals and the refinement of every subject, the method of traditional literature retrieval, because of lack of links between document content, can hardly satisfy people's need. Co-word analysis is a document content analysis method, which can reflect the relationship of the co-words by analyzing the words appeared together in the same document, and then reveals the relationship of different subjects of one certain field which represents by the documents. By co-word analysis, researchers in different fields can find out the hot research points; they can analyze the relationship of different subjects and research in different levels, they can also provide extended information retrieval service. Many researchers have improved this method and also worked on it both in theory and application in many fields. Currently, the application research mainly focuses on information retrieval, information science and information system, scientometrics, artificial intelligence and so on.This paper firstly researches on the basic principle of co-word analysis; inquires into the basic procedures and technology of this method; introduces the software tools and system platform that commonly used; analyzes the current problems in co-word analysis such as ignoring the different contributions of documents in different types and keywords with different importance; and based on these problems proposes a weighted co-word analysis method including vertical weighted, horizontal weighted and mixed weighted.The basic idea of weighted co-word analysis is that by attaching each document itself and each keyword in the documents with different weight to identify the importance of the document or keywords, this makes the documents with high quality and the improtant keywords play leading function. Bibexcel and SPSS are selected as the co-word analysis tools in the experiment, and the original experimental data are periodicals with selected subject from CNKI. The experimental results before and after weighted show that the weighted co-word analysis method can reflect different research purposes, thus making the analysis result more efficiency; it decreases the sensitivity of co-word analysis and makes the subject orientations expressed by the clustering cloud more clear and accurate.