Open Source Summit Latin America 2022

In a world increasingly connected, speed is prioritized over truth, in a way that information have a considerable power of reaching many people in a few time, and the truth of such information is questioned and, maybe proven, after a long period of digital dissemination. In such way, disinformation can be disastrous for society in general and valuable for those who benefit from it. With so much data, could it be possible to have machine learning algorithms helping us detect information that have a high chance of being false? Could this process be accomplished in a ethical way? In this presentation João Lucas will discuss such questions, and present the results of applying the Passive Aggressive algorithm to create models that classify as true and fake Brazilian news. Moreover, he will talk about the patterns that he found in those news, using knowledge extraction techniques (i.e. t-SNE), and how those patterns helps to identify features that could be relevant to build even more precise models.