1. Consider the following product review sample from Amazon. helpful is a variable indicating the total number of helpful votes each review has received so far; score refers to the rating of the product being reviewed. Suppose your goal is to use the review text to train a model for predicting the sales of a product (which is highly correlated with its average rating on Amazon)

(1) [6 points] Which one of the two terms, flavor or buck, is more informative about review #4 in the review corpus? Justify your answer by calculating their TF-IDF scores.

Note: The TF-IDF scores should be calculated after word-stemming and removing stopwords and non-words. Assume that stopwords = c(is, and, the, of) and word-stemming replaces nouns in their plural form with their singular form. Please specify your own list of non-words.

(2) [2 points] In this analysis, should we pre-process the review data by removing numbers from the text? Why or why not? Please explain.