It's just a simple classifier based on this Kaggle dataset. This code was written on Kaggle. You can upload it to Kaggle and run it there or download the dataset and run it on local machine. FastText was used as the word to vector embedding layer and LSTM was our deep learning model.
- It will take about 5 hours to run the model. So, hold on.
- I ran this model on Kaggle. At first it ran for 1 hour and got disconnected. On second try, it went till epoch 2 (2 hours) and then it disconnected. After wasting 3 hours, I gave up and didn't run it again. After epoch 2, accuracy was 85%. It is a simple beginner level model. This could be optimized in so many ways to improve performance or avoid overfitting. I will update it over time when I learn better technique.