An analysis of structure of an epic: A statistical approach*

Abstract

This study establishes the feasibility of applying some of the well-known statistical methods for analysing the structure of Cilappathikaaram, one of the five epics of Tamil literature. There are three divisions in this text. The first division contains 10 chapters. The second division contains 13 chapters and the third division contains seven chapters. For this analysis all 30 chapters are considered, and it is a complete enumerative study. A complete concordance is formed using the metric form of this epic, facilitating an accurate compilation of statistics such as character and word frequencies. It is found that the frequencies of occurrence of characters in the three divisions are not from the same distribution and it is also determined that the three divisions are mutually different. The two entropy measurements are obtained for each of the three divisions based on the estimated unigram probabilities and from the estimated word-length distributions. These estimates suggest that the entropy estimate is lowest for division two and highest for division three. The word-length frequency curves of the three divisions are unimodal, positively skewed, and leptokurtic. But the word-length frequency distribution of the third division is significantly different from that of the other two divisions. The existence of natural clusters is established by grouping all thirty chapters using clustering techniques on the basis of different statistics of their word-length distributions.