test 3 different RNN using GloVe embedding to classify sentiment of tweets
Dataset and GloVe embedding file were too big to load into GitHub
dataset: https://www.kaggle.com/datasets/kazanova/sentiment140/
GloVe embedding: glove.6B.50d.txt from https://nlp.stanford.edu/projects/glove/
Using GloVe embedding reduces number of trainable parameters by 2.5 million.
Each model had around 80% accuracy on testing dataset.