Student: Uriel Merola Minage e Silva
Title: Métodos de detecção de fake news: Comparativo entre abordagens crowd signals e métodos de comitê
Advisors: Jorge de Abreu Soares (Advisor) and Ronaldo Ribeiro Goldschmidt (Co-advisor)
Committee: Jorge de Abreu Soares (Cefet/RJ), Ronaldo Ribeiro Goldschmidt (IME), Eduardo Bezerra da Silva (Cefet/RJ), Paulo Marcio Souza Freire (IME)
Day/Time: September 14, 2023 / 10:30a.m.
Room: Cefet/RJ – Unidade Maracanã, Bloco E, sala E-520
Abstract: The significant rise of fake news dissemination is due mainly to the easy generation and consumption of information provided by social networks. Several machine learning-based approaches have been proposed to detect and combat this malicious kind of information. Among the leading approaches to detect fake news there is one based on hybrid crowd signals (HCS). To identify false information, this approach combines signals (i.e. opinions on whether the information is false or not) collected either from the users of social networks or from machine learning classifiers. Although promising, as far as we could observe, HCS employs a naive bayes classifier (i.e. a naive method) to combine the signals from the crowd and infer which pieces of news are false. Hence, the present work raises the hypothesis that ensemble methods applied to combine the opinions provided by the machine learning classifiers used in HCS and implicit opinions provided by users in social networks
can lead to better classification models. The experiments conducted in this study provide evidence of the validity of the raised hypothesis.