LAC: Latent Authorial Clustering of shorter texts

Abstract

Authorial clustering of a corpus of documents results in grouping them so that each group encircles those written by the same author. LAC, the presented software, implements a novel approach towards authorial clustering of paragraph-long texts. It operates on documents modelled in low-dimensional languageindependent topic-based vector spaces via a flexible command-line interface. Assessing the goodness of such representations and comparing them to other approaches are made easier with LAC. In addition, LAC facilitates studying semi-supervision for authorial clustering and how to best determine the number of authorial clusters 𝑘 in a corpus.