The Fictionality of Non-Fiction: A Computational Study of the Heroic Reportage in the People’s Liberation Army Daily, 1956-1989

Inspired by Andrew Piper’s computational study of fictionality, this project develops a machine model (TF-IDF classifier) to automatically distinguish between fictional and non-fictional texts, using twentieth-century Chinese novels and articles from Chinese Wikipedia, respectively, as the training dataset (54,000 samples). Achieving 97.1% accuracy on the test set (6,000 samples), the model is then applied to analyze more than 185,000 articles published between 1956 and 1989 in the People’s Liberation Army Daily, the official journal of China's armed forces that has been instrumental in shaping the revolutionary narrative in the PRC. The journal is not only fully digitized but also continued publication throughout the Cultural Revolution (1966-1976), constituting an unparalleled historical resource. The computational analysis yields a paradoxical result: although the genre of “fiction” has been marginalized in the journal during the CR years, the ostensibly informative, non-novelistic texts published in the late 1960s become increasingly fictional, rising from the 50% average in 1956 to more than 65% in 1969. This finding raises important questions in the light of Richard Walsh’s theory of fictionality, which posits that fictionality of a text is determined by the reader's interpretive assumption provoked by extra-diegetic context surrounding the text, not its diegetic content. The author argues that the ambiguous status of the PLA Daily articles situated the Cultural Revolution hero within the historical reality of post-1949 China (the global informative frame of the news report), while inviting the reader's poetic-affective engagement through novelistic rhetoric (the local narrative frame).