This project is one of the getting started competitions in kaggle.The aim of this contest is to classify whether tweets are disaster or not.
This repo includes data exploration, data preprocessing and model creation stages.I used 2 different models. One of them is Glove with LSTM and the other is Bert model.I used keras and ktrain libraries in the model creation phase.The data preprocessing stage can be improved for higher accuracy
The Glove with LSTM model accuracy roughly: 0.81 The Bert Model: 0.84
You can reach the competition from this link:https://www.kaggle.com/c/nlp-getting-started Also you can find embeddings from this link:https://www.kaggle.com/thanakomsn/glove6b300dtxt