# Create flags, percentage score, and bins for reviews$helpful
reviews$helpful.nan = sapply(reviews$helpful,
                             function(x) ifelse(x[2]==0, 1,
                                                ifelse(x[1]>x[2], 1, 0)))
reviews$helpful.perc = sapply(reviews$helpful,
                              function(x) ifelse(x[2]==0, NA,
                                                 ifelse(x[1]>x[2], NA, x[1]/x[2])))
reviews$helpful.bins = cut(reviews$helpful.perc,
                           breaks = 3,
                           include.lowest = T,
                           labels = c("Lower", "Middle", "Upper"))

# Subset reviews for EDA, text mining, and modeling
reviews.mod = reviews[reviews$helpful.nan==0, ]

#count number of words per review
reviews.mod$numWords <- sapply(gregexpr("\\W+", reviews.mod$reviewText), length)

#correlation of Helpful percentage and depth of review
cor(reviews.mod$helpful.perc, reviews.mod$numWords)