Um modelo de verificação de fatos baseado em crowdsourcing

Abstract

The process of creating content on the Internet was immensely promoted by the emergence of Social Networks, and with this, the volume of information kept on the web grows enormously. If, on the one hand, the dissemination of information has become facilitated by these technologies, on the other hand, ensuring the veracity of this information has become more difficult, since social networks often allow the publication of content without moderation. Thus, social networks today represent a great potential for disseminating false news. This fact prompted the emergence of factchecking organizations, usually represented by news organizations. News organizations, in turn, have their action limited to their productive capacity, which may not meet the great demand that arises from the immense volume of fake news, to which we are submitted in this era of misinformation. This dissertation then proposes a crowdsourcing-supported fact-checking model, which is capable to filter, analyze and classify news, proposing an alternative, proportional to the pace of fake news creation. In addition to the model proposal, the work comprises the development of a tool to support the process operation, involving the crowd work management. To evaluate the model, experiments were performed using Amazon Mechanical Turk, and at the end, the results were discussed, where the potential of the crowds use in fact-checking tasks was observed in the context of the proposed model.