This goal of this project is to create a classifier that can detect political sentiment. The current classifier uses BERT as the underlying algorithm and only classifies between two categories(Liberal and Conservative). This classifier has a 71% accuracy and we hope to improve it by getting new data from different forms online.
As part of the project, I learned about various algorithms including LSTMs and RNNs. I also learned how to train and test these large scale classifiers efficiently in Tensorflow and PyTorch. The end goal of the project is to turn the classifier into a browser plug in that could be used on any site.