This seems like the most unintuitive approach teaching a brand new concept (Naive Bayes). You've just given up a quick example of how to train a readily formatted dataset, and before you know it, it's the next step where you're throwing something complicated and expecting us to fill in the ellipsis in your code template. No explanation of how it's supposed to be done.
Calculate P(positive) and P(negative)? Sure. How about tell me how it's done first instead of referring us to the mathematical formula?