Authorship Attribution, Idiolectal Style, and Online Identity: a specialised corpus of Najdi Arabic tweets

This thesis investigates a synergy of approaches, namely corpus linguistics, stylistics, computational linguistics, and computer-mediated communication to address forensic authorship attribution problems in Arabic. It aims to explore how authorship as a concrete matter can communicate and reveal the theoretical notion of the idiolect. On another front computer sciences, specifically in machine learning, tackle authorship from a standpoint of accuracy. Corpus linguistics is a common ground where both fields meet, as linguists and computer scientists use corpora to test their methods. Linguists approach it from a stylistic, qualitative point of view because it provides them with the explicability a courtroom would require when presenting their analysis report as expert witnesses. Computational scientists, on the other hand, focus on the quantitative, statistical aspect of analysis which generated, until recently, black box tools that do not necessarily show the tool’s trail of analysis.