I have a dataframe containing 50+ features of items (all normalized and/or encoded). These also include the ratings and sentiment polarity of comments left by users for the item
