Deep representation takes the sequence of embedding vectors as input and converts them to a compressed representation. The compressed representation effectively captures all the information in the sequence of words in the text. The deep representation part is usually an RNN, though a CNN can be utilized as well. For the RNN, the compressed representation of the text is the final hidden state of the network output after the last time step. In the CNN, this will be the output of the last layer, which would commonly be a max pooling layer. Both the RNN and CNN outputs, therefore, capture deep representations of the text inputs.