Uma abordagem em cascata para predição de faixa etária de autores a partir dos seus textos

Abstract

With the widespread use of the Internet, text mining tools have become important for companies' business strategies. At the same time, the growing volume of anonymous texts available on the Internet poses a challenge to obtaining useful information, for example, in the development of personalization models for digital commerce and marketing, as well as in fields such as digital forensics. In the area of Authorial Characterization — an area that seeks to predict social and demographic characteristics of a group based on their texts —, although there are a large number of studies, they present inequalities regarding the characteristics analyzed, the language used and the classification approaches employed. Thus, this work seeks to propose and evaluate a cascade classification method, which combines traditional classification methods with a dictionaries-based model, in order to predict author age ranges based on texts written in Portuguese. The results found show that the dictionaries-based model in combination with traditional models can bring improvements in age range classification.