Look through the prediction_dataset.ipynb notebook to see how to get the Amazon data.
You can then generate a pickle file which you use in the other notebooks.
The Predicting spelling errors.ipynb notebook has the code to try and find spelling errors in the Amazon data.
The Predicting grammar errors.ipynb notebook has all the steps needed to see whether a sentence contains any grammatical errors.
The final notebook, Predicting paraphrases.ipynb shows how to predict whether some of the sentences are paraphrases off eachother.