Abstract
In order to apply Self-Organizing Map (SOM) for text of involves the wide-ranging field keywords, a system of clustering and visualization by using Wikipedia category is proposed. Wikipedia's category has a hierarchical structure that it can uniquely identify the word and can be add flexibility in additional learning.
The proposed system firstly generates input vectors from Wikipedia category in relation to keywords. In this method, input vectors are generated from Wikipedia article keywords taken from each text and the corresponding Wikipedia category.
In consideration of the hierarchical structure in Wikipedia category, input vector is created based on keywords in common category. Using the proposed method, the problem of dynamic changes in texts and reconfiguration of vector elements can be solved. Moreover, information loss in newly obtained keyword can be prevented. The obtained map represents the nature of relations of the laws relation nature which would be intuitively understandable to the users. To confirm the effectiveness of the proposed system, evaluation experiments measured the classification accuracy and computational time using 400 news data.
