Statistical analysis of Chinese classic text structure of keywords in commentaries of Huang-ti Yin-fu-ching

Abstract

In a previous paper the authors considered a statistical method for analyzing Chinese classic text by computer and applied them to a commentary of Huang-ti Yin-fu-ching. This method is to make a permutated matrix by a two-way clustering which shows the mutual relationship between important words and commentary units, which are assumed to be units of meaning in the commentary. In spite of its usefulness, this method has following problems to be overcome. (1) It is difficult to compare different permutated matrices from different commentaries. (2) The previous method depends upon commentary units which is a particular form of commentary. Therefore its application is severely restricted.In the present paper new methods are considered and discussed to solve these two problems. For problem (1), we try cluster analysis of commentaries by different authors based on unified re-phrasing of the original canon and the commentary to enable comparison of different permutated matrices. For problem (2), instead of a commentary unit, we consider a sentence and neighborhood of a keyword as a unit of meaning. These methods can be applied Chinese classic texts of any format. In particular the latter method need not any manual phrasing.