Methods for improving food-related personalization

Patent metadata

Field	Value
Publication number	US-11587140-B2
Application number	US-202016803550-A
Country	US
Kind code	B2
Filing date	Feb 27, 2020
Priority date	Jan 25, 2018
Publication date	Feb 21, 2023
Grant date	Feb 21, 2023

Abstract

Systems and methods for improving food-related personalization for a user including generating a recipe database including a set of recipe data structures; deriving a recipe vector representation of the recipe data structures; determining a set of user food preferences; extracting a set of recipe vector constraints from the set of user food preferences; determining a personalized food plan for the user, including automatically selecting a subset of the set of recipe data structures associated with recipe vector representations that satisfy the set of recipe vector constraints; determining fulfillment parameters for grocery items associated with the personalized food plan; and automatically facilitating fulfillment of grocery items associated with the personalized food plan based on the fulfillment parameters.

First claim

We claim:

1. A method for improving food-related personalization for a user, comprising:
determining a set of recipe data structures from a set of recipes, comprising: parsing each recipe into a preparation parameter set and an ingredient entity set, identifying references to ingredient entities of the ingredient entity set in the preparation parameters of the preparation parameter set, and mapping the preparation parameters and the respective ingredient entities together, wherein a recipe data structure is modified with a substitute instruction based on the mapping;
deriving a recipe vector representation of each of the set of recipe data structures, comprising using a trained neural network model to determine the recipe vector representation comprising values equivalent to a set of weights of an intermediate layer of the trained neural network model;
deriving an ingredient vector representation associated with each ingredient entity of the set of ingredient entities using the trained neural network model;
determining a set of substitution parameters associated with each ingredient entity of the set of ingredient entities based on the ingredient vector representation, comprising comparing the ingredient vector representation of each ingredient entity with the ingredient vector representation of each other ingredient entity;
determining a set of user food preferences; and
determining a personalized food plan based on the recipe vector representation of each of the set of recipe data structures and the set of user food preferences.

2. The method of claim 1 , wherein the set of recipe data structures comprises natural language data.

3. The method of claim 1 , wherein the neural network model is trained using semi-supervised learning.

4. The method of claim 1 , wherein the set of user food preferences are determined at least in part based on social media content.

5. The method of claim 1 , wherein the set of user food preferences comprises a user allergy.

6. The method of claim 1 , further comprising determining a constraint associated with the set of user food preferences, wherein determining the personalized food plan comprises:
comparing the recipe vector representation of each of the set of recipe data structures to the constraint; and
selecting recipe data structures of the set of recipe data structures that satisfy the constraint based on the comparison.

7. The method of claim 6 , wherein the constraint is an item availability constraint.

8. The method of claim 1 , further comprising automatically facilitating fulfillment of grocery items associated with the personalized food plan.

9. The method of claim 8 , wherein automatically facilitating fulfillment of grocery items comprises facilitating physical transport of the grocery items.

10. The method of claim 1 , further comprising determining fulfilled items, and updating the personalized food plan based on the fulfilled items.

11. A method for improving food-related personalization for a user, comprising:
determining a recipe data structure based on recipe data, comprising: parsing the recipe data into a preparation parameter set and an ingredient entity set, identifying references to ingredient entities of the ingredient entity set in the preparation parameters of the preparation parameter set, and mapping the preparation parameters to the respective ingredient entities;
deriving a recipe vector representation of the recipe data structure, including using a trained neural network model to determine the recipe vector representation having values equivalent to a set of weights of an intermediate layer of the trained neural network model;
determining a recipe vector constraint associated with the recipe data structure;
determining a first ingredient vector representation of a first ingredient entity of the set of ingredient entities and a second ingredient vector representation of a second ingredient entity of the set of ingredient entities, using the trained neural network model;
determining that the second ingredient entity can be substituted for the first ingredient including: comparing the second ingredient vector representation to the recipe vector constraint and based on the comparison, determining that the second ingredient vector satisfies the recipe vector constraint; and
generating a personalized food plan based on a comparison between the recipe vector constraint and the recipe vector representation.

12. The method of claim 11 , wherein the trained neural network model comprises a plurality of neuronal layers and wherein the recipe vector representation comprises an intermediate layer of the plurality of neuronal layers.

13. The method of claim 11 , wherein the recipe data structure is associated with the set of ingredient entities of a plurality of ingredient entities.

14. The method of claim 11 , further comprising modifying a preparation parameter of the set of preparation parameters based on a modification to the set of ingredient entities associated with the recipe data structure.

15. The method of claim 14 , wherein a connected cooking device is controlled according to the preparation parameter subsequent to modifying the preparation parameter.

16. A method for improving food-related personalization for a user, comprising:
determining a set of recipe data structures from a set of recipes, comprising: parsing each recipe into a preparation parameter set and an ingredient entity set, identifying references to ingredient entities of the ingredient entity set in the preparation parameters of the preparation parameter set, and mapping the preparation parameters and the respective ingredient entities together, wherein a recipe data structure is modified with a substitute instruction based on the mapping;
deriving a recipe vector representation of each of the set of recipe data structures, including using a trained neural network model to determine the recipe vector representation including values equivalent to a set of weights of an intermediate layer of the trained neural network model;
selecting a substitute ingredient entity for an ingredient entity of the ingredient entity set based on a probability that a food professional would agree that the substitute ingredient entity is an admissible substitute for the ingredient entity; wherein the probability is calculated based on a ranking of substitute ingredients by food professionals;
determining a set of user food preferences; and
determining a personalized food plan based on the recipe vector representation of each of the set of recipe data structures and the set of user food preferences.

Assignees

• Kraft Foods Group Brands Llc

Inventors

• Leifer Tjarko
• Andrejko Erik
• Aldor-Noiman Sivan

Classifications

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0895

  Weakly supervised learning, e.g. semi-supervised or self-supervised learning · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06F16/90324

  using system suggestions · CPC title