R Deep Learning Essentials - Second Edition

Comparing the deep learning NLP architectures

Here is a summary of all of the models in this chapter, ordered by their sequence in this chapter. We can see that the best traditional machine learning approach got 95.24%, which was beaten by many of the deep learning approaches. While the incremental changes from the best traditional machine learning to the best deep learning model may seem small at 0.80%, it reduces our misclassified examples by 17%, which is a significant relative change:

Model	Accuracy
Best traditional machine learning approach	95.24%
Simple deep learning approach	94.97%
1D convolutional neural network model	95.73%
Recurrent neural network model	93.90%
Long short term memory model	95.37% ...