Distance measures for exploring pairs of novels in a large corpus of Italian literature

In text clustering most distance-based methods summarize the occurrences of a set of linguistic features to obtain a distance. It should decrease when texts are written by the same author, however, there are further properties that might influence the result: gender of the authors, their age, their geographical origin, publication date of the novels, their size, etc. In this study, regression analyses compare the performance of three distances and highlight, among available covariates, the preeminent effect of the author's hand but also interesting patterns in the effect of novels’ size.