Identifying substitute ingredients using a natural language processing system

Patent metadata

Field	Value
Publication number	US-9519620-B1
Application number	US-201615009905-A
Country	US
Kind code	B1
Filing date	Jan 29, 2016
Priority date	Jan 29, 2016
Publication date	Dec 13, 2016
Grant date	Dec 13, 2016

Abstract

A first food ingredient may be received. The first food ingredient may comprise a first plurality of chemical constituents. A plurality of candidate food ingredient substitutes may be received. Each candidate food ingredient substitute may comprise a second plurality of chemical constituents. For each of the plurality of candidate food ingredient substitutes, a quantity of the second plurality of chemical constituents that match the first plurality of chemical constituents may be determined. One or more food ingredient substitutes may be identified based on at least the quantity of the second plurality of chemical constituents that match the first plurality of chemical constituents.

First claim

What is claimed is:

1. A computer-implemented method for identifying substitutes for ingredients in a recipe based on analyzing chemical association, the method comprising:
generating, by a natural language processing system, a second set of new recipes based on at least a machine learning associated with a first set of preexisting recipes, wherein the machine learning includes identifying patterns and associations between a first set of particular ingredients of the first set of preexisting recipes and identifying patterns and associations between cooking preparation steps of the first set of preexisting recipes and based on the patterns and associations between the first set of particular ingredients and further based on the patterns and associations between the cooking preparation steps, generating the second set of new recipes;
in response to the generating of the second set of new recipes, receiving, by the NLP system, a first food ingredient, the first food ingredient comprising a first plurality of chemical constituents;
receiving, by the NLP system, a plurality of candidate food ingredient substitutes, each candidate food ingredient substitute comprising a second plurality of chemical constituents;
determining, by the NLP system and for each of the plurality of candidate food ingredient substitutes, a quantity of the second plurality of chemical constituents that match the first plurality of chemical constituents; and
identifying, by the NLP system, one or more food ingredient substitutes based on at least the quantity of the second plurality of chemical constituents that match the first plurality of chemical constituents.

2. The method of claim 1 , further comprising:
ranking each of the plurality of candidate food ingredient substitutes based on the determining; and
generating, in response to the ranking, a first set of recipes, wherein each recipe of the first set includes the one or more food ingredient substitutes.

3. The method of claim 1 , wherein the plurality of candidate food ingredient substitutes includes a second food ingredient and a third food ingredient, and wherein the determining includes:
determining that the second food ingredient shares a higher quantity of chemical molecules with the first food ingredient than the third food ingredient shares with the first food ingredient, wherein the second food ingredient sharing a higher quantity of chemical molecules with the first food ingredient corresponds to the second food ingredient having a higher percentage of intersecting chemical molecules with the first food ingredient than the third food ingredient does; and
ranking the second food ingredient as a higher candidate for substitution than the third food ingredient based on the second food ingredient sharing a higher quantity of chemical molecules with the first food ingredient than the third food ingredient shares with the first food ingredient, wherein the one or more food ingredient substitutes includes the second food ingredient but not the third food ingredient.

4. The method of claim 1 , further comprising grouping the first food ingredient into a particular food type class, wherein the receiving of the plurality of candidate food ingredient substitutes is based on the particular food type class that the first food ingredient is in.

5. The method of claim 1 , further comprising:
ranking each of the plurality of candidate food ingredient substitutes based on at least comparing a first aroma description of each of the plurality of candidate food ingredient substitutes with a second aroma description of the first food ingredient; and
generating, in response to at least the ranking, a first set of recipes that includes the one or more food ingredient substitutes.

6. The method of claim 1 , wherein the identifying the one or more food ingredient substitutes is further based on:
determining a third plurality of chemical constituents of the first food ingredient as it exists before the first food ingredient is cooked;
determining a fourth plurality of chemical constituents of the first food ingredient as it exists after the first food ingredient is cooked;
comparing the third plurality of chemical constituents with the fourth plurality of chemical constituents; and
determining, based on the comparing, that the fourth plurality of chemical constituents differs from the third plurality of chemical constituents above a threshold, wherein the first plurality of chemical constituents is the fourth plurality of chemical constituents.

Assignees

• IBM

Inventors

• Pinel Florian
• Seabolt Edward E

Classifications

• G06F40/00 Primary

  Handling natural language data (speech analysis or synthesis, speech recognition G10L) · CPC title

• G06F40/30

  Semantic analysis · CPC title

• G06F16/313

  Selection or weighting of terms for indexing · CPC title

• G06F16/34

  Browsing; Visualisation therefor (browsing or visualisation for clustering or classification G06F16/358) · CPC title

• G06F40/295

  Named entity recognition · CPC title