Agu, Emmanuel O.
BOS / Boston, MA Project Center
Fake news is increasingly pervasive, and we address its problematic aspects to help people intelligently consume news. In this project, we research machine learning models to extract objective sentences, encouraging unbiased discussions based on facts. The most accurate model, a convolutional neural network, achieves an accuracy of 85.69%. The team implemented an end-to-end web system that highlights objective sentences in user input to make our model publicly accessible. The system also provides additional information about user input, such as links to related web pages. We evaluate our system both qualitatively by interviewing users, and quantitatively with surveys consisting of rating scale questions. Received positive feedback indicates the usability of our platform.
Worcester Polytechnic Institute
Major Qualifying Project
All authors have granted to WPI a nonexclusive royalty-free license to distribute copies of the work, subject to other agreements. Copyright is held by the author or authors, with all rights reserved, unless otherwise noted.