Simulating the client messages data

Simulating text messages that actually make sense is very hard, and we won't attempt it here. Instead, what we'll do is leverage a dataset that was published about food reviews on Amazon. The dataset was published as part of the paper published by McAuley and Leskovec, From amateurs to connoisseurs: modeling the evolution of user expertise through online reviews, 2013. You can find the dataset in Kaggle (https://www.kaggle.com/snap/amazon-fine-food-reviews). We won't show the code that prepared the data for this example, but basically, what it does is rename the variables we want STARS, SUMMARY, and MESSAGE, delete the rest, and save the data frame into the reviews.csv file. For the interested reader, the code that accomplishes this task, as well as the original and processed data, is inside the code repository for this book ...