This includes Network Models for Emojifier: Standard RNNs and LSTM.
In order to get the program to work apart from getting prediction by loading in pre-trained model, we need to have pre-trained GloVE embedding vectors. It could be downloaded from one of the current research in Stanford: https://nlp.stanford.edu/projects/glove/
Before using the file make_prediction.py, it is noted that the file will load a saved model from emojifier_LSTM.h5, so therefore, without the emojifier_LSTM.h5, it is not able to make any prediction.
The model is located in the file named emojifier_basic.py. It is vital to note that this basic model has no saving capability. Everytime you want to get a prediction for a sentence, you have to go append new sentence into sentences list variable in the file and the model will re-train itself.
The model is located in the file named emojify_LSTM.py. Since Keras library has been used for this model, it can save all the parameters into a file which is emojfier_LSTM.h5. Now, if you re-train the model by using LSTM, the saved model would be overwritten by the new one.
You can go into make_prediction.py. In the list_of_sentences list variable, feel free to delete and append new items so as to test out the application.