Automatic Extraction of Structures of Issues Based on Newspaper Articles

摘要

In this article, we propose the method to extract the structure of issues by using newspaper articles. The principal method used to extract the structure is ISM (Imperative Structural Modelling). ISM is widely used for structuring problems. However, using ISM often consumes much time and efforts. This is mainly because building the relation matrix, which is an essential element for building ISM diagram, needs time and cost. In this article, we focus on the automatic relation matrix generation algorithm. Firstly, keywords related to a issues are selected. Then the relation between two arbitrary keywords is examined. After examining the relations of all pairs of keywords, a relation matrix is generated from which a reachable matrix is calculated. By using this reachable matrix, an ISM diagram can be constructed. Then, the correlation ratio such as co-occurrence ratio of all words pair will be calculated and the average correlation ratio is determined. By comparing each correlation ratio and the average ratio, the strong and week correlation will be determined. After the determination of all pairs of correlation, relation matrix will be generated automatically. The proposed algorithm has been implemented as an Web application.