A text as unique as a fingerprint: Text analysis and authorship recognition in a Virtual Learning Environment of the Unified Health System in Brazil

Abstract

Authorship Attribution, the act of deducing the author of a given text based on its writing characteristics, is an issue with an extensive history. It refers to the task of properly recognizing the text’s author within a specific group of candidates, based on relevant features extracted from the text (Stylometry). Hence, stylometry identifies relevant attributes that define a space in which authors can be distinguished. Because writers use language in different ways to express their ideas, linguistic variations make it possible to recognize authorship. The definition of the author’s text is discussed, in this article, as an auxiliary tool in the distance education platform of the Ministry of Health, AVASUS. Therefore, the stylometric features were extracted from the collected data set, and different classification algorithms were trained. The objective was to predict the authorship of texts with more than 30 characters. As a result, an acknowledgment text for the Virtual Learning Environment of the Unified Health System in Brazil was obtained. The precision achieved in the classification process was 92% in some classifiers. This aspect suggests that techniques for extracting stylometric features may be used to recognize the author of a given text.