Stylometry of literary papyri

Abstract

In this paper we present the first results of stylometric analysis of literary papyri. Specifically we perform a range of tests for unsupervised clustering of authors. We scrutinise both the best classic distance-based methods as well as the state-of-the-art network community detection techniqes. We report on obstacles concerning highly non-uniform distributions of text size and authorial samples combined with sparse feature space. We also note how clustering performance depends on regularisation of spelling by means of querying relevant annotations.