TartanHacks 2017
Members: John Curcio, Jeffrey Yan
Title: Polictical Party Voice Classifier
Here is a hack that classifies audio as coming from a liberal or conservative talk show host.
Using an LSTM neural network, we are able to train a model to differentiate and classify short audio clips as either InfoWar's Alex Jones or The Daily Show's Trevor Noah. We took long .WAV files and divided them into very small samples to use as our training data. These data were then fed into the neural network which created optimal weights. The result is a model that was able to classify new audio clips with 97% accuracy.
The long-term goal is to extend this model with more talk show hosts to be able to classify any voice recording as either conservative or liberal politically motivated. With the great divide that has formed in the United States recently, it is now more important than every to understand and empathize with beliefs and thoughts that conflict with our own. This hack aims to be a small stepping stone in closing the wide gap that has separated so many people by learning what it means to be red or blue and then teaching that to others.
Log in or sign up for Devpost to join the conversation.