M.S. AAI Capstone Chronicles 2024

Detecting Fake News Using Natural Language Processing The figure above depicts a webpage constructed using Gradio, facilitating the input of news articles. Upon clicking the submit button, a pre-trained model is invoked to predict whether the news is fake or real. In addition to the prediction, the webpage displays explanations highlighting keywords, distinguishing between fake and true news. These explanations are presented both visually in an image format and in downloadable HTML format, enhancing the user's understanding of the prediction process. Conclusion Future Directions In the future, we aim to incorporate multimodal (text and images) prediction into our system, a secondary goal that couldn't be explored in the current project scope. One prominent multimodal dataset, Fakeddit, features Reddit-like posts with text and image pairs, comments, and metadata, boasting over a million samples (Nakamura et al., 2020). While the dataset's size offers learning advantages, it also poses challenges due to significant time and computational requirements, potentially requiring sampling for balance. For image data, a Convolutional Neural Network (CNN) architecture is preferable, such as custom models or pre-trained options like ResNet or VGG-16, which would complement the LSTM model in tandem. Model Productionization In real-world implementation, user interaction with the model is pivotal. Hosting the model on an online interface or mobile app, either standalone or integrated into social media platforms, is proposed for single-instance prediction with minimal latency. Given the evolving nature of information, regular data updates and model retraining are imperative, facilitated by a Continuous Integration/Continuous Deployment (CI/CD) pipeline. Monitors will be employed to

19

44