In recent decades, the media that we consume as a population has become more and more politically polarized. This increasing divide may be due to a number of factors, but one significant cause is the way that news organizations, social media, and other online platforms tailor the content they serve to an individual based on their previous interests. These “filter bubbles” can make it very difficult to acquire one’s news from a balanced and diverse set of news publications. Neural networks are a fascinating and increasingly important field of study in computer science, and there have been significant advances in the field in the past couple of decades. This paper delves into the process of creating a neural network trained on a collection of 4,000 news articles from eight publications, for the purpose of analyzing and classifying future news articles according to their political biases. This neural network would allow for a web-interface or mobile application that users could use to recognize the biases in the news they are reading, strive to diversify their news sources, and stay well-informed in a politically balanced way. Of the neural network variations explored (a standard LSTM, a model consisting of both a CNN and an LSTM, and a bidirectional LSTM), the bidirectional LSTM was found to achieve the greatest accuracy with the least overfitting.