Granted Patent B2

US 12,333,303 · App. 18/427,591 · Granted Jun 17, 2025

Inventors: Oliver Zahn (Berkeley, CA); Karthik Sekar (Berkeley, CA); Richard Gerkin (Berkeley, CA)

Assignee: Climax Foods Inc.

G06F9/06G01N33/0001G01N33/02G06N20/00G16C20/70

Abstract

Claims (22)

1. A method, comprising:

determining a composition for a sample;

determining a context for the sample, wherein the context comprises at least one of a sample phase or a sample substrate;

learning a flavor function using flavor training data collected from set of sensory panelists; and

determining a flavor characterization for the sample based on the flavor function, the context, and the composition.

2. The method of claim 1 , further comprising training a mixture model using the flavor training data, wherein the sample comprises a mixture of sample components, wherein learning the flavor function comprises learning a component flavor function for each sample component, and wherein determining the flavor characterization for the sample comprises:

determining a component flavor characterization for each sample component based on the component flavor function and the composition; and

using the mixture model, aggregating the component flavor characterizations based on the context to determine the flavor characterization for the sample.

3. The method of claim 2 , wherein the flavor training data comprises panelist flavor rankings for single-component samples and panelist flavor rankings for multi-component samples, wherein the component flavor function for each sample component is learned using the panelist flavor rankings for single-component samples, wherein the mixture model is trained using the panelist flavor rankings for multi-component samples.

4. The method of claim 1 , further comprising:

determining a context for the sample; and

parameterizing the context, wherein learning the flavor function comprises training a flavor model to determine the flavor function based on the composition and the parameterized context.

5. The method of claim 1 , wherein determining the composition comprises predicting the composition using a trained composition model.

6. The method of claim 5 , wherein the composition comprises a gustation composition, wherein the gustation composition is predicted using the trained composition model based on at least one of: headspace composition, manufacturing process parameters, manufacturing ingredients, molecular structures of sample components, or sample matrix.

7. The method of claim 5 , wherein the composition comprises an olfactory receptor composition, wherein the olfactory receptor composition is predicted using the trained composition model based on at least one of: headspace composition, manufacturing process parameters, manufacturing ingredients, molecular structures of sample components, or sample matrix.

8. The method of claim 1 , wherein the flavor function comprises a sigmoid function relating composition to component flavor intensity.

9. The method of claim 8 , wherein learning the flavor function comprises learning an offset, a slope, and a maximum value for the sigmoid function.

10. The method of claim 1 , wherein the flavor training data comprises panelist flavor rankings.

11. The method of claim 10 , wherein learning the flavor function comprises training a flavor model to determine the flavor function based on the composition, wherein training the flavor model comprises:

predicting a flavor characterization for each of a set of training samples using the flavor model;

determining a predicted flavor ranking for the set of training samples based on the predicted flavor characterizations; and

training the flavor model based on a comparison between the predicted flavor ranking and the panelist flavor ranking.

Assignments (1)

Continuity (5)

References Cited (26)