Ebola-Category-Predictions

Description

To predict the category to which a Tweet on Ebola belongs to from data mined from Twitter. This was done as part of my Summer Internship at CNeRG, IITKGP.

Neural Networks Architectures used:

• LSTM
• Stacked LSTM
• Bidirectional LSTM
• Stacked Bidirectional LSTM ( For all these the 200D twitter GloVe embeddings of Standford NLP was used)
• Multi Layer Perceptron ( For this Universal Sentence Encoder was used instead of GloVe embeddings)

Results

The maximum test accuracy of 65% was achieved in Stacked Bidirectional LSTM followed by 60% in case of Multi Layer Perceptron using Universal Sentence Encoder.