Diachronic delta: A computational method for analysing periods of accelerated change in literary datasets

Abstract

Abstract In his recently published book Distant Horizons (2019), Ted Underwood argues that literary change takes place incrementally. By analysing a dataset of English language fiction obtained from the digital research library HathiTrust, Underwood demonstrates that from the early 18th to the 21st century, literature written in English evinces a steady increase in words associated with physical description, concrete nouns as well as sensory perception and that this increase is accompanied by a proportional decrease in words associated with politics, economics as well as religion (Underwood 25). This article represents an effort to test Underwood’s incremental hypothesis by operationalizing Peter W.H. Smith and W. Aldridge’s cosine distance-based emendation to John Burrows’ ‘Delta’ method (Smith, P. W. H. and Aldridge, W. (2011). Improving authorship attribution: optimising burrows’ delta method. Journal of Quantitative Linguistics 18(1): 63–88). By applying cosine Delta across time rather than as it is conventionally applied, across text (Evert, Stefan et al. Understanding and Explaining Delta Measures for Authorship Attribution. Digital Scholarship in the Humanities 32 (suppl_2) (2017): ii4–ii16), we can identify particular years as being associated with extensive amounts of ‘novelty’ (which years introduce the most amount of distance from their predecessors) and also possess extensive amounts of ‘resonance’ (are relatively proximate to their successors). According to Alexander T.J. Barron et al.’s method for analysing the parliamentary debates of the French revolution, agents within a dataset which have high amounts of novelty and high amounts of resonance are both innovative and influential, as they are significantly different from the years which come before and relatively similar to the years which come after. We might refer to these years as ‘breaks’ and contrast their behaviour with the longue-durée approach otherwise prevalent within the computational literary studies discourse. Though, as this article demonstrates, Underwood’s hypothesis of incremental change remains robust when considering the changes which come over the novel from the mid-18th to the early 20th century, applying this method to poetic and dramatic production over the same period of time, thereby examining the distinct ways in which novelty and resonance behave, allow us to modulate Underwood’s hypothesis and to understand literary history within a more conjunctural framework.