Language processing

So far, we have seen how we can train a simple feedforward neural network on Keras for an image classification task. We also saw how we can mathematically represent image data as a high-dimensional geometric shape, namely a tensor. We saw that a higher-order tensor is simply composed of tensors of a smaller order. Pixels group up to represent an image, which in turn group up to represent an entire dataset. Essentially, whenever we want to employ the learning mechanism of neural networks, we have a way to represent our training data as a tensor. But what about language? How can we represent human thought, with all of its intricacies, as we do through language? You guessed it—we will use numbers once again. We will simply translate our texts, which are composed of sentences, which themselves are composed of words, into the universal language of mathematics...