The goal is to introduce a model that would be capable of generating new recipes based on a user's input ingredients.
Basing on Eat-PIM and FlavorGraph generated embeddings, the input user's ingredients would result in a FlowGraph, presenting the recipe proposed by the model.
- Parse provided ingredients
- Match infringements with Foodon and Wikidata entities
- Choose most relevant ingridients
- Generate a set of the most applicable actions for these ingredients
- Create a FlowGraph from these ingredients and actions
- Generate textual recipe based on the FlowGraph