Work

Beyond Standard Language: Detecting Rule-Changing Creativity in Experimental Literature.

Year: 2026
Type: article
Abstract: NLP tools are predominantly designed for high-resource, standardised languages, and comparable limitations affect language varieties within a single language — diatopic, chronological and sociolinguistic varieties, as well as the idiolectal varieties deliberately constructed by authors for literary purposes. In experimental literary texts, deviations from the standard are not discardable noise but triggers of significance, and their computational treatment requires methodologies that preserve a high degree of interpretability. The paper presents two complementary approaches to the detection of rule-changing creativity. The first, bottom-up, builds regular expressions modelled on deviation patterns previously identified through close reading, filtering matches against standard lexical resources; it has been applied to two corpora of twentieth-century Italian experimental authors, Giovanni Testori and Stefano D'Arrigo. The second, top-down, applies information-theoretic measures (surprisal and entropy) to a corpus of eleven twentieth-century Italian novels, testing whether high surprisal peaks in low-entropy contexts can function as detectors of linguistic creativity, with POS distribution used to enrich interpretability. Their combination outlines a virtuous circle in which quantitative markedness detection and qualitative stylistic analysis mutually refine one another.
Language: English
Open Access status: green
OpenAlex ID: W7204201006
FWCI:
Cites:
Cited by:
Related to: 10