This is an abstractive text summarizer on Amazon food reviews data set. This is done using encoder-decoder seq2seq LSTM model.

1. Run test.ipynb to calculate the model weight.
2. Alternatively you can simply run test_predict.py to predict the summary.
3. You can also load trainedModel_weights3.h5 and use it. It has max_text_len = 30 and max_summary_len = 8, trained over 41368 and tested over 4581 data set with accuracy rate of 66.60%.
4. Initially trainedModel_weights2.h5 is loaded and used. It has max_text_len = 30 and max_summary_len = 10, trained over 42241 and tested over 4679 data set with accuracy rate of 72.19%.
5. You can also load trainedModel_weights.h5 and use it. It has max_text_len = 50 and max_summary_len = 10, trained over 59683 and tested over 6613 data set with accuracy rate of 70.56%.
6. Total data set is above 500K but only 100K is loaded to train and test the model.