Creating a word-to-token dictionary for inference

During preprocessing, we trained a Keras tokenizer to replace the words with their numerical word indices, so that the processed movie reviews could be fed to the LSTM model for training. We have also kept the first 50000 words with the highest word frequency, and have set the review sequences to be of a maximum length of 1000. Although the trained Keras tokenizer was saved for inference, it cannot be used by the Android app directly. We can restore the Keras tokenizer and save the first 50000 words and their corresponding word indices in a text file. This text file can be used in the Android app, in order to build a word-to-indices dictionary to convert the words of the review text to their word indices. It is important to note that the word to indices mapping can be retrieved from the loaded Keras tokenizer object, by referring...