Real Time Fallacy Detection
Disciplines
Artificial Intelligence and Robotics | Computer Sciences
Abstract (300 words maximum)
In today's fast-moving era, social media often creates echo chambers where users see content that supports their existing beliefs, whether the arguments are sound or not. This combined with shrinking attention spans, makes it difficult for users to critically assess the information they consume. The objective of this project is to assist with this problem by creating a tool for the automated detection of logical fallacies, providing a much-needed counterbalance against flawed reasoning and helping to enhance digital literacy.
Our project's core idea is a three-stage process. First, an audio or video file of spoken arguments is transcribed into text. Second, this text is cleaned and prepared for analysis. Finally, a classifier model performs multi-class classification on the text to identify and label a specific fallacy.
To build our classifier, we will implement and compare several supervised learning models, such as Naive Bayes, Logistic Regression, and Support Vector Machine (SVM). We will use a combination of datasets for training, including the LOGIC dataset for its diverse collection of fallacies and the Propaganda Detection dataset. Combining these sources will help us train a more robust model. We will measure the success of our project using standard metrics like accuracy, weighted precision, recall, and weighted F1-Score to evaluate the model's overall performance
Use of AI Disclaimer
no
Academic department under which the project should be listed
CCSE – Computer Science
Primary Investigator (PI) Name
Md Abdullah Al Hafiz Khan
Real Time Fallacy Detection
In today's fast-moving era, social media often creates echo chambers where users see content that supports their existing beliefs, whether the arguments are sound or not. This combined with shrinking attention spans, makes it difficult for users to critically assess the information they consume. The objective of this project is to assist with this problem by creating a tool for the automated detection of logical fallacies, providing a much-needed counterbalance against flawed reasoning and helping to enhance digital literacy.
Our project's core idea is a three-stage process. First, an audio or video file of spoken arguments is transcribed into text. Second, this text is cleaned and prepared for analysis. Finally, a classifier model performs multi-class classification on the text to identify and label a specific fallacy.
To build our classifier, we will implement and compare several supervised learning models, such as Naive Bayes, Logistic Regression, and Support Vector Machine (SVM). We will use a combination of datasets for training, including the LOGIC dataset for its diverse collection of fallacies and the Propaganda Detection dataset. Combining these sources will help us train a more robust model. We will measure the success of our project using standard metrics like accuracy, weighted precision, recall, and weighted F1-Score to evaluate the model's overall performance