by Ryan Walker •

CincyHackathon

My project: peanut_butter_jelly_time

I’ve decide to help Meruni on their challenge to develop a way to discover complementary products. For example, “jelly” goes with “peanut butter”.

Challenge #1: Data

I was hoping that Meruni’s 2011 program-mate Receept would have some grocery shopping receipt data to leverage, but they don’t have any – woulda been cool to mash them up. So then I had to look for some other source of data, Meruni recommended looking at recipe data.

They like Punchfork, but to get recipe data requires paid API, and they had no way for us to upgrade free account via web. Oops, lost sale there, PunchFork.

I’ll use Recipe Puppy and the Google Prediction API.

Here’s the process:

  • Pull a list of popular recipes.
  • Search RecipePuppy to get a list of ingredients for each popular recipe.
  • Create training data in CSV to upload to Google Prediction API.
  • Build simple web app where one can type an ingredient in, and get top 10 ingredients that complement it.

Wish me luck!

Update