Dominique: An AI-Powered Fact-Checking Chatbot for Democratizing Access to Reliable Information

Abstract

Access to information has become easier than ever. However, this convenience has introduced a significant challenge: the spread of fake news. This phenomenon has become a global problem, posing a major threat to sectors such as politics, education, and health. Consequently, the ability to distinguish reliable from misleading or harmful sources, known as fact-checking, has become a very necessary skill. Despite several studies on detecting fake news, few reach the public and aim to promote digital literacy. Another potential limitation is that most studies focus solely on training performance, without addressing the challenges posed by real-world testing. To overcome these limitations, we developed Dominique, a user-friendly Brazilian Portuguese chatbot tool trained with a group of Machine Learning models and five embedding techniques. The proposed tool was evaluated on previously unseen news data. Moreover, a new Brazilian Portuguese dataset comprising over 20,000 news articles was created. The main goal of the tool development is to provide an accessible solution and, through news content provided by a user, report the probability of the information being fake, encouraging the user to verify the accuracy of the information, draw their own conclusions, and become better informed.