F-Detector: Design of a Solution Based on Machine Learning to Detect Logical Fallacias on Digital Texts

The design of F-Detector, as an improved Machine Learning-based solution aimed at detecting logical fallacies in digital texts. Using an advanced NLP approach based on BERT, enhances the classification of text fragments into various categories of logical fallacies by incorporating more robust linguistic features and a more accurate classification model. The performance of F-Detector was rigorously evaluated using multiple metrics, demonstrating an overall precision of $75 \%$, significantly improving upon the $18 \%$ precision achieved by previous solutions. Additionally, this model will be integrated into a Chrome extension, allowing users, including educators, researchers, and the general public, to utilize it in a natural way. This enhanced tool has the potential for broad application in education and digital content analysis, contributing to better discourse quality and more effective misinformation detection.