A stylometric analysis of newspapers, periodicals and news scripts

Abstract

Differences in the writing style of authors provide the grounds for authorship attribution or classification studies. This paper attempts to quantify the literary style of various forms of media, including broadsheet and tabloid newspapers, technical periodicals and television news scripts. The aim is to investigate the richness of vocabulary exhibited in these texts under the proposition that the writing style usually varies depending on the targeted readership or audience. In this paper we show the importance of using maximum likelihood estimates for the three parameters of the Sichel distribution, as opposed to using the inverse Gaussian Poisson distribution, which is a particular case. We then use multivariate pattern recognition techniques such as discriminant analysis, classification trees and neural networks to establish differences in the afore-mentioned types of media.