Uma Estratégia para Moderação de um Grande Conjunto de Comentários de Usuários.

Abstract

The expansion of social media and the advent of Web 2.0 promoted the participation of persons interested in exposing their opinions on what it intends to discuss a collective environment or on any facts reported by the press . However, due to legal mechanisms to exert control over a particularly offensive punch, with expressions that attack personalities material, it becomes of great interest to the classification of documents relating to comments entered by users of news sites, in order to identify which may or may not be disclosed in the digital environment, avoiding judicial providers demands of these environments. This work proposes the use of automatic classification techniques to identifying reviews the disclosure in the media should be allowed or not, aiding humans in the work of comment moderation. For both, various techniques in data processing, such as reducing words to their canonical form, dimensionality reduction and weighting terms were explored. All these techniques have been studied in order to model an algorithm able to mimic human decisions to release or not the comment.