Um texto tão singular quanto a impressão digital: o uso de sistemas inteligentes para reconhecimento de autoria

Abstract

The Authorship Attribution, the science of inferring an author for a particular text based on their writing characteristics, is a problem with a long history. In this work, it is being proposed the study of the problem of attribution and recognition of authorship in order to make it a tool of use in the distance learning platform of the Ministry of Health (MS), AVASUS. The techniques of text analysis and stylometric characteristics of the authors will be presented, allowing the authorship to be determined in significantly better indexes, in which the texts are greater than 30 characters. This proposal targets AVASUS, where students take the courses of the platform, share their interests and thoughts in the form of messages in the forums and perform activities that require writing on certain topics in the health area. These written productions are the focus of the application of attribution and recognition of authorship. The techniques studied as a proposal are a two-stage process, where in the first stage, stylometric information is extracted from the collected data set and in the second stage, different classification algorithms are trained and lexical, semantic or syntactic analysis techniques are applied to predict the authors of the texts. The effort is to maximize the accuracy of predictions with optimal amount of data and users under consideration.