Good Things Come in Pairs: A Computational Study of Poetic Parallelism in the Six Dynasties

This article outlines the computational methodology for a multipart study on poetic parallelism in the Six Dynasties (222–589), a formative era in the history of Chinese literature. As the first part of this project, the authors develop a quantitative-formalist definition of parallelism, synthesizing insights from modern linguistics, vector semantics, and traditional critics such as Liu Xie 劉勰 (ca. 465–522) and Zhong Rong 鍾嶸 (ca. 468–518), alongside key passages from Kūkai's 空海 (774–835) Discourse on the Secret Treasury of Literary Mirrors (Bunkyō hifuron 文鏡秘府論). They then evaluate several statistical methods for automatic detection of parallelism in classical Chinese verse, including BERT-based classifiers and few-shot prompting of generative models such as GPT-4o and DeepSeek R1. Applying these techniques to over twenty-five thousand pentasyllabic couplets from Lu Qinli's 逯欽立 (1910–1973) Poetry of the Pre-Qin, Han, Wei, Jin, and Northern and Southern Dynasties (Xian Qin Han Wei Jin Nanbeichao shi 先秦漢魏晉南北朝詩), the study traces the macroscopic evolution of parallelism across five periods: Jian'an (建安, 196–220), Zhengshi (正始, 240–249), Taikang (太康, 280–290), Yuanjia (元嘉, 424–453), and Yongming (永明, 483–493). The key finding is the demonstrable increase in the usage and sophistication of parallel structures over time, a rise in complexity that can be understood as a growing number of dimensions along which poets constructed and pursued comparison. This confirms the Six Dynasties as a pivotal era for this poetic technique and validates the quantitative approach for future studies of the five individual periods and their representative poets.