Best Parameters When Using Stylo on Romanian Literary Texts Cover Image

Best Parameters When Using Stylo on Romanian Literary Texts
Best Parameters When Using Stylo on Romanian Literary Texts

Author(s): Maria-Corina Dimitriu, Lucreția Pascariu, Roxana Patras
Subject(s): Computational linguistics
Published by: Editura Tracus Arte
Keywords: natural language processing (NLP); stylometry in R; computational stylistics; stylo package; authorship attribution

Summary/Abstract: AbstractAuthorship attribution is one of the fields that has benefited most from the recent developments in stylometry. However, there are no general guidelines on the parameters to be used for reliable authorship attribution, since the performance of stylometric analysis is highly influenced by the language and length of the texts, as well as the composition of the corpus. This paper aims to test the performance of the stylo package on two Romanian corpora of short texts, one of them homogeneous and the other heterogeneous. The overarching objective of our research is to propose some guidelines for authorship attribution tasks on Romanian short literary texts and see to what extent the corpus composition influences the results. Likewise, by comparing the results obtained with the intuitive stylo package to those obtained with considerably more sophisticated digital tools, we aim to point to the on-the-shelf feasibility of reliable stylometric analysis even in less-resourced languages.

  • Issue Year: XXI/2025
  • Issue No: 2 (42)
  • Page Range: 201-216
  • Page Count: 16
  • Language: English
Toggle Accessibility Mode