Fractality in Chinese prose

Abstract

Abstract This study explores long-range correlations in terms of sentence or segment length variation in Chinese narrative texts and nonfiction prose. (Drożdż et al. 2016, Quantifying origin and character of long-range correlations in narrative texts. Information Sciences, 331 32–44) analyzed Western novels and found fractal patterns, defined as self-similar, wavelet recurrence, and alternation. Inspired by this study, our research tries to determine whether similar patterns commonly exist in Chinese literature and compares the similarities and differences with Western literature. We calculated the Hurst exponent, β-values, and Δα values for ninety-five Chinese novels, ranging historically from late Qing to contemporary Internet novels, covering the geopolitical regions from Mainland China to Taiwan and Hong Kong. We also made comparisons with pre-modern vernacular novels, historical texts in classical Chinese, contemporary nonfiction and expository writings, as well as randomly generated texts. We found that Chinese novels exhibit fractal patterns as well. In particular, the texts exhibit a better fractal quality if the sentence lengths are measured by Chinese characters, instead of words. There is no clear correlation between fractality and cultural–political contexts and individual authors, but historically speaking, modern Chinese texts show stronger long-range correlations than pre-modern texts. Moreover, long-range correlations in Chinese literature are weaker than in Western literature, and there is a lower percentage of novels with multifractal structures. Our data also show that the fractality not only exists in literary texts, but also in nonliterary, non-narrative, and expository writings; yet, there is no long-range correlation in randomly generated texts. We further conclude that fractality is a fundamental feature of prose writing and human writing.