• Null models in authorship analysis - an alternative approach to established methods in stylometry

    Author(s):
    Steffen Pielström (see profile)
    Date:
    2019
    Group(s):
    DH2020
    Subject(s):
    Authorship, Language and languages--Style--Statistical methods, Literary style--Statistical methods
    Item Type:
    Conference paper
    Conf. Title:
    DH2020
    Conf. Org.:
    ADHO
    Conf. Loc.:
    https://hcommons.org/
    Conf. Date:
    July 20.-25. 2020
    Tag(s):
    delta, hypothesis testing, null models, Authorship attribution, Computational stylistics, Stylometry
    Permanent URL:
    http://dx.doi.org/10.17613/j6qj-nw10
    Abstract:
    In stylometry, authorship questions are currently approached in terms of two separate tasks: authorship attribution and authorship verification. Recent research in stylometry often interpreted both as classification tasks and concentrated on further improving the accuracy of these two procedures. This study proposes an alternative approach to authorship questions, treating them as hypothesis tests based on an empirical null model. The central question this approach evolves around: what does a text distance value of X actually mean for my authorship question? This essay outlines how a null model can be derived from empirical observations to answer this question. The approach allows to choose a rejection criterion for the null hypothesis that two texts have been written by different people, producing reasonable estimations for the alpha and beta error.
    Notes:
    Abstract to the presentation shared under https://hcommons.org/deposits/objects/hc:31810/datastreams/CONTENT/content#frame6315
    Metadata:
    Status:
    Published
    Last Updated:
    2 years ago
    License:
    Attribution
    Share this:

    Downloads

    Item Name: pdf pielstroem2020-null_models_in_authorship_analysis.pdf
      Download View in browser
    Activity: Downloads: 84