Multifractal analysis of Chinese literary and web novels

Abstract

Multifractal analysis of literary works, utilizing measures such as sentence length and word length as time series, have revealed the prevalence of multifractal structures within these works. After constructing a corpus of 712 Chinese novels, we performed multifractal detrended fluctuation analysis (MF-DFA) and multifractal detrending moving average analysis (MF-DMA) on them. Both method revealed that non-monotonic generalized Hurst exponential curves and distorted spectral functions appeared in about half of the web novels, but were rarely found in literary novels. By conducting a meticulous examination of sentence length sequences and adjustment of the time series range covered by the computation, it becomes apparent that this irregular Hurst exponent curve stems from the uneven distribution of long sentences within the work. This uneven distribution mirrors the instability present in the author’s creative style, which is intimately linked to the characteristics of the publishing mechanisms for web novels.