Granted Patent B2

US 12,327,275 · App. 17/118,279 · Granted Jun 10, 2025

Inventors: Chul Lee (San Francisco, CA); Kyler Eastman (Austin, TX); Patrick Howell (Austin, TX); Layla Martin (San Francisco, CA); Hesamoddin Salehian (San Francisco, CA); Joohyun Kim (San Francisco, CA)

Assignee: MyFitnessPal, Inc.

G06Q30/0631G16H10/60G16H20/60

Abstract

Claims (57)

1. A method for operating a health tracking system, the method comprising:

receiving a crowd-sourced data record comprising at least a descriptive string and nutritional data regarding a consumable item to which the data record corresponds;

determining one of a plurality of possible tastes associated to the consumable item by:

evaluating the nutritional data of the data record to determine an accuracy of the nutritional data with respect to the consumable item,

when the evaluation of the nutritional data is determined to be accurate, applying a first statistical model to the descriptive string to determine a first set of probabilities that the consumable item has each of the plurality of possible tastes, wherein the plurality of possible tastes includes a plurality of fundamental flavors, applying a second statistical model to the nutritional data to determine a second set of probabilities that the consumable item has each of the plurality of possible tastes, and determining the one of the plurality of possible tastes for the consumable item based on the first set of probabilities and the second set of probabilities, wherein applying the first statistical model comprises (i) vectorizing the descriptive string, (ii) comparing the vectorized descriptive string to a set of training vectors, and (iii) using a weighted k-nearest neighbor algorithm to calculate the first set of probabilities that the consumable item has each of the plurality of possible tastes based on a similarity between the vectorized descriptive string and vectors in the set of training vectors, and

when the nutritional data is determined to not be accurate, determining the one of the plurality of possible tastes by applying the first statistical model to the descriptive string to determine the first set of probabilities that the consumable item has each of the plurality of possible tastes and determining the one of a plurality of possible tastes for the consumable item based only on the first set of probabilities;

updating a crowd-sourced database in order to associate the determined taste with the data record in the crowd-sourced database;

performing the acts of receiving, determining, and associating updating with respect to each of a plurality data records stored in the crowd-sourced database such that each of the plurality of data records is associated to a taste;

determining a taste profile specific to a user, said taste defined by a probability of each of the fundamental flavors;

generating a list of recommended ones of a plurality of consumable items for the user based at least in part on the determined taste profile specific to the user and the tastes associated with the plurality of data records; and

transmitting the generated list of recommended ones of a plurality of consumable items to a health tracking device associated with the user.

2. The method according to claim 1 , wherein the nutritional data includes at least a total caloric content and respective amounts of a plurality of macronutrients of the consumable item, and the act of determining the accuracy further comprises comparing the total caloric content to a caloric content representative of the respective amounts of the plurality of macronutrients.

3. The method according to claim 1 , further comprising: when the nutritional data is determined to be accurate, weighting an output of the first statistical model, and weighting an output of the second statistical model.

4. The method according to claim 3 , wherein a value by which the output of the first statistical model is weighted and a value by which the output of the second statistical model is weighted are dependent on a value of the output of the first statistical model.

5. The method according to claim 1 , further comprising:

receiving a list of consumable items, the received list being associated with a user profile stored in the database, the received list containing a plurality of entries corresponding to individual ones of the data records selected by the user; and

associating the determined taste profile to the user profile in the database, wherein the determined taste profile specific to the user profile is based on one or more patterns determined from the tastes associated to each of the data records corresponding to the plurality of entries in the received list.

6. The method according to claim 5 , further comprising:

selecting individual ones of a plurality of advertisements stored in the database based on the determined taste profile; and

transmitting the selected individual ones of the plurality of advertisements to a health tracking device associated with the user profile.

7. The method of claim 1 wherein the plurality of fundamental flavors include two or more of sweet, salty, umami, sour, spicy and bitter.

8. The method of claim 1 wherein the taste profile is defined by a user probability for each of the fundamental flavors.

9. A health tracking system comprising:

a database configured to store a plurality of crowd-sourced data records, each of the plurality of data records comprising at least a descriptive string and nutritional data regarding the consumable item to which the data record corresponds; and

a data processor in communication with the database, the data processor being configured to (i) receive one or more of the plurality of data records from the database, (ii) determine a taste aspect for the consumable item to which each of the received ones of the one or more of the plurality of data records corresponds, the determination being based on an evaluation of at least one of the descriptive string and the nutritional data, wherein the taste aspect defines probabilities for each of a plurality of fundamental flavors (iii) generate a list of recommended ones of the plurality of data records based at least in part on the determined taste aspects corresponding to the plurality of data records, and (iv) send the generated list of recommended ones of the plurality of data records to a health tracking device associated with the user;

wherein the evaluation of at least one of the descriptive string and the nutritional data comprises:

for each of the one or more of the plurality of data records, determine whether the nutritional data thereof is accurate with respect to the consumable item;

when it is determined that the nutritional data is accurate, apply a first mathematical model to the nutritional data to calculate a probability that respective ones of the consumable item have a first taste aspect and based on the result thereof, apply a second mathematical model to the descriptive string, wherein applying the second mathematical model comprises (i) vectorizing the descriptive string, (ii) comparing the vectorized descriptive string to a set of training vectors, and (iii) using a weighted k-nearest neighbor algorithm to calculate a first set of probabilities that the consumable item has each of the plurality of possible tastes based on a similarity between the vectorized descriptive string and vectors in the set of training vectors; and

when it is determined that the nutritional data is not accurate, apply the second mathematical model to the descriptive string to calculate a probability that respective ones of the consumable item have the first taste aspect.

10. The health tracking system of claim 9 , wherein the data processor is further configured to store the determined taste aspects in the database in association with the respective ones of the one or more of the plurality of data records.

11. The health tracking system of claim 9 , wherein the at least one of the first and second mathematical models model also comprises a machine learning model which utilizes a training set of data records having known taste aspects associated therewith.

12. The health tracking system of claim 9 , wherein applying the first mathematical model also comprises using a weighted k-nearest neighbor algorithm.

13. The health tracking system of claim 9 , wherein the application of the second mathematical model to the descriptive string further comprises

vectorize the descriptive string; and

apply applying a pattern recognition model to the vectorized descriptive string.

14. The health tracking system of claim 13 , wherein the nutritional data includes at least a total caloric content and respective amounts of a plurality of macronutrients of the consumable item, and wherein determining whether the nutritional data thereof is accurate with respect to the consumable item further comprises comparing the total caloric content to a caloric content representative of respective amounts of the plurality of macronutrients.

15. The health tracking system of claim 13 , wherein when the nutritional data is determined to be accurate, weighting an output of the first mathematical model, and weighting an output of the second mathematical model.

16. The health tracking system of claim 15 , wherein a value by which the output of the first mathematical model is weighted and a value by which the output of the second mathematical model is weighted are dependent on a value of the output of the first mathematical model.

17. The health tracking system of claim 9 , wherein the data processor is further configured to:

receive a list of consumable items, the received list being associated with a user profile stored in the database, the received list containing a plurality of entries corresponding to individual ones of the plurality of data records; and

associate the determined taste aspect to the user profile in the database, wherein the determined taste aspect specific to the user profile is based on one or more patterns determined from the tastes associated to each of the data records corresponding to the plurality of entries in the received list.

18. The health tracking system of claim 9 , wherein the data processor is further configured to:

select individual ones of a plurality of advertisements stored in the database based on the user profile; and

transmit the selected individual ones of the plurality of advertisements to a health tracking device associated with the user profile.

19. A non-transitory computer-readable medium for operating a health tracking system, the computer-readable medium having a plurality of instructions stored thereon that, when executed by a processor, cause the processor to:

receive a plurality of crowd-sourced data records from a database, the plurality data records each comprising a descriptive string and nutritional data regarding the respective corresponding consumable;

determine a taste for each of the respective corresponding consumables of the plurality of data records based on at least one of (i) the descriptive string of the data record, and (ii) evaluation of the nutritional data of the data record, the nutritional data including at least a total caloric content and respective amounts of a plurality of macronutrients of the consumable, wherein the determined taste is defined by probabilities that the consumable has each of a plurality of fundamental flavors;

wherein the evaluation of the nutritional data of the data record comprises determining the accuracy of the nutritional data with respect to the consumable by comparing the total caloric content to a caloric content representative of the respective amounts of the plurality of macronutrients, wherein,

when the nutritional data is determined to be accurate, determining the taste by (a) applying a first statistical model to the descriptive string and (b) applying a second statistical model to the nutritional data, wherein applying the first statistical model comprises (i) vectorizing the descriptive string, (ii) comparing the vectorized descriptive string to a set of training vectors, and (iii) using a weighted k-nearest neighbor algorithm to calculate the first set of probabilities that the consumable item has each of the plurality of possible tastes based on a similarity between the vectorized descriptive string and vectors in the set of training vectors; and

when the nutritional data is determined to be inaccurate, determining the taste by applying the first statistical model to the descriptive string and omitting to utilize the nutritional data;

update the database by storing the determined tastes in the database associated to the respective ones of the plurality of data records;

derive a taste profile for a particular user; and

recommend one or more items for consumption and/or provide one or more advertisements to said particular user based at least in part on the derived taste profile and the determined taste for each of the respective corresponding consumables.

20. The non-transitory computer-readable medium of claim 19 , wherein the plurality of instructions are further configured to:

utilize said tastes for each of said corresponding consumables to derive the taste profile for the particular user, the taste profile being based at least in part on one or more of:

individual ones of the plurality of data records selected by the particular user, an age of the particular user, and/or

a gender of the particular user.

Assignments (4)

Continuity (2)

References Cited (44)