Throughout this chapter, we've discussed various NLP techniques, particularly with regards to preprocessing text. In many use cases, we will need to interact with an ANN to perform the final analysis. The type of analysis is not relevant to this section, but imagine you're developing a sentiment analysis ANN. You appropriately tokenize and stem your training text, then, as you attempt to train your ANN on your preprocessed text, you realize you have no idea how to get words into a neural network.
The simplest approach is to map each input neuron in the network to an individual unique word. When processing a document, you can set the input neuron's value to the term frequency (or absolute count) of that word ...
