I’m trying to use hadoop to get some statistics from a json file like average number of stars for a category or language with most reviews. To do this I am using mrjob, I found this code: It allows to find the most used word, but I am not sure how to do this with json attributes instead of words.