Computational stylometry: predicting the authorship of investment arbitration awards

Abstract

This chapter presents an overview of the legal stylometry literature, and its utility in providing insight into the authorship of otherwise anonymous legal texts. After first exploring the use of machine learning stylometry models, and their relevance to analysing the texts produced by investor-state arbitration proceedings, the chapter then demonstrates how the stylistic features of these texts can be used to infer authorship. The resulting analysis suggests both that there is significant variation in styles between academic writing and the applied legal writing that arises in arbitration proceedings, and that there is significant variety within arbitral award writing styles, even when the texts have the same authorship.