Next Word Prediction Using LSTMs

This is in continuation with the main article Next Word Prediction using Swiftkey Data . I will discuss in detail the text features and architecture of the LSTM models over the following sections . I have created sequences of different lengths from the tokens of the cleaned corpus maintaining the order of the tokens . I have created sequences of length 2, 4 and 7 . I have encoded the words in the sequences using Keras Tokenizer and splitted the sequences such that only the last word in the sequence is in y and the rest is in x . Thus a sequence of length 2 will have first word in x and the last in y .