Patent Issued for Token based communications for machine learning systems (USPTO 11847246): United Services Automobile Association

2024 JAN 10 (NewsRx) -- By a News Reporter-Staff News Editor at Insurance Daily News -- A patent by the inventors Gout, Michael P. (San Antonio, TX, US), Rodgers, Fredrick William (San Antonio, TX, US), filed on September 12, 2018, was published online on December 19, 2023, according to news reporting originating from Alexandria, Virginia, by NewsRx correspondents.

Patent number 11847246 is assigned to United Services Automobile Association (San Antonio, Texas, United States).

The following quote was obtained by the news editors from the background information supplied by the inventors: “Personally identifiable information, or sensitive personal information is information that can be used on its own or with other information to identify, contact, or locate a single person, or to identify an individual in context. In general, communicating personally identifiable information between entities (such as corporations) is subject to regulatory restrictions. Further, information that an organization knows about an individual (such as a customer), and internal business practices are generally considered proprietary information with business value.”

In addition to the background information obtained for this patent, NewsRx journalists also obtained the inventors’ summary information for this patent: “In general, an innovative aspect of the subject matter described in this specification can be embodied in methods that include the act of generating a token representative of private data. The method includes the act of identifying at least one entity associated with the private data. The method includes the act of associating the token with at least one entity. The method includes the act of communicating the token and associated entity as label data for training of a machine learning system. The method includes the act of associating the token and associated entity with additional corresponding data as feature data for training of a machine learning system. The method includes the act of training a machine learning system using the combined label and feature data. The method also includes the act of using the trained machine learning system to make one or more inferences on entities for the token.

“In general, another innovative aspect of the subject matter described in this specification can be embodied in methods that include act of generating a token representative of private data. The methods include the act of identifying at least one entity associated with the private data. The methods include the act of associating the token with at least one entity. The methods also include the act of providing information identifying at least one entity and the token to a machine learning system. The methods also include the act of receiving information from a trained machine learning system on entities.

“In general, another innovative aspect of the subject matter described in this specification can be embodied in methods that include act of receiving a list of entities, each entity associated with at least one token. The methods include the act of obtaining additional information from the machine learning system associated with the corresponding entity. The methods include the act of generating training data including training examples, each training example including the additional information associated with an entity of at least some entities and a token from the at least one token associated with the entity. The methods also include the act of training a machine learning system using the training data.

“In general, another innovative aspect of the subject matter described in this specification can be embodied in methods that include the act of receiving from a trained learning machine information identifying one or more entities likely to be associated with a token, where the trained learning machine is trained to make inferences about entities based on a token that is indicative of private data and does not contain information sufficient for the operator of the trained learning machine to identify the private data.

“Implementations can optionally include one or more of the following features, alone or in combination. At least one entity may be identified by an identifier shared with the machine learning system. The token may be representative of an event that has occurred. The token may be representative of membership in a group. The token may not contain information sufficient to identify the private data. The methods may include the acts of providing one or more individual identifiers to a machine learning system and receiving additional information from the machine learning system identifying at least some of the one or more entities likely to be associated with the token.”

The claims supplied by the inventors are:

“1. A computer-implemented method performed by at least one processor, the method comprising: generating, by at least one processor, for each of a plurality of events, a respective token that represents the event, wherein the respective token representing the event is encrypted using a private key associated with an organization and allows the organization to identify the event but does not allow a third party to identify the event, wherein the plurality of events comprises at least one of (i) one or more life events comprising at least one of a marriage, a divorce, a childbirth, or a job change, or (ii) one or more major purchases comprising at least one of a purchase of a house, a purchase of a car, or a purchase of a vacation package; for each event of the plurality of events: identifying, by at least one processor, at least one entity associated with the event; associating the respective token encrypted by the private key of the organization and representing the event with the at least one entity; communicating, by at least one processor, the respective token and the at least one entity as label data for training of a machine learning system; and associating the respective token encrypted by the private key and representing the event and the at least one entity with behavioral data of the at least one entity to generate feature data for training of the machine learning system; combining the label data and the feature data to generate training data for training the machine learning system; training the machine learning system on the training data that is a combination of label data and feature data generated for the plurality of events; and using the trained machine learning system to make one or more inferences on entities for the respective tokens, comprising: for each of one or more events in the plurality of events, identifying, using the trained machine learning system, a set of new entities that each having a probability of experiencing the event within a predetermined time frame, wherein each probability is greater than zero, and filtering, by at least one processor, from the set of new entities, entities for which the probability of experiencing the event is below a threshold to obtain an updated set of new entities; and providing, over a computer network, information on the updated set of new entities to the organization.

“2. The computer-implemented method of claim 1, wherein the at least one entity has an identifier shared with the machine learning system.

“3. A computer-implemented method performed by at least one processor, the method comprising: generating, by at least one processor, for each of a plurality of events, a respective token that represents the event, wherein the respective token representing the event is encrypted using a private key associated with an organization and allows the organization to identify the event but does not allow a third party to identify the event, wherein the plurality of events comprises at least one of (i) one or more life events comprising at least one of a marriage, a divorce, a childbirth, or a job change, or (ii) one or more major purchases comprising at least one of a purchase of a house, a purchase of a car, or a purchase of a vacation package; for each event of the plurality of events: identifying, by at least one processor, at least one entity associated with the event; associating the respective token encrypted by the private key of the organization and representing the event with the at least one entity; and providing information identifying the at least one entity and the respective token to a machine learning system for training the machine learning system; using the trained machine learning system to make one or more inferences on entities for the respective tokens, comprising: for each event of one or more events in the plurality of events, identifying, using the trained machine learning system, a set of new entities that each having a probability of experiencing the same event within a predetermined time frame, wherein each probability is greater than zero, and filtering, by at least one processor, from the set of new entities, entities for which the probability of experiencing the same event is below a threshold to obtain an updated set of new entities; and providing, over a computer network, information on the updated set of new entities to the organization.

“4. The computer-implemented method of claim 3, wherein the at least one entity has an identifier shared with the machine learning system.

“5. A computer-implemented method comprising: receiving, by at least one processor, from a trained learning machine, information identifying one or more entities likely to be associated with a token, wherein the token represents an event of a plurality of events, wherein the token representing the event is encrypted using a private key associated with an organization and allows the organization to identify the event but does not allow a third party to identify the event; wherein the plurality of events comprises at least one of (i) one or more life events comprising at least one of a marriage, a divorce, a childbirth, or a job change, or (ii) one or more major purchases comprising at least one of a purchase of a house, a purchase of a car, or a purchase of a vacation package, wherein the trained learning machine is trained to make inferences about entities based on the token that is indicative of the event and does not contain information sufficient for an operator of the trained learning machine to identify the event, wherein the trained learning machine is configured to: for each event of one or more events in the plurality of events, identify a set of new entities that each having a probability of experiencing the same event within a predetermined time frame, wherein each probability is greater than zero, and filter, from the set of new entities, entities for which the probability of experiencing the same event is below a threshold to obtain an updated set of new entities; and provide, over a computer network, information identifying the updated set of new entities.

“6. A non-transitory computer storage medium storing instructions which, when executed by at least one processor, cause the at least one processor to perform operations comprising: generating, for each of a plurality of events, a respective token that represents the event, wherein the respective token representing the event is encrypted using a private key associated with an organization and allows the organization to identify the event but does not allow a third party to identify the event, wherein the plurality of events comprises at least one of (i) one or more life events comprising at least one of a marriage, a divorce, a childbirth, or a job change, or (ii) one or more major purchases comprising at least one of a purchase of a house, a purchase of a car, or a purchase of a vacation package; for each event of the plurality of events: identifying at least one entity associated with the event; associating the respective token encrypted by the private key of the organization and representing the event with the at least one entity; communicating the respective token and associated entity as label data for training of a machine learning system; and associating the respective token encrypted by the private key and representing the event and the associated entity with behavioral data of the event to generate feature data for training of the machine learning system; combining the label data and the feature data to generate training data for training the machine learning system; training the machine learning system on the training data that is a combination of label data and feature data generated for the plurality of events; and using the trained machine learning system to make one or more inferences on entities for the respective tokens, comprising: for each of one or more events in the plurality of events, identifying, using the trained machine learning system, a set of new entities that each having a probability of experiencing the same event within a predetermined time frame, wherein each probability is greater than zero, and filtering, from the set of new entities, entities for which the probability of experiencing the same event is below a threshold to obtain an updated set of new entities; and providing, over a computer network, information on the updated set of new entities to the organization.

“7. The non-transitory computer storage medium of claim 6, wherein the at least one entity has an identifier shared with the machine learning system.”

There are additional claims. Please visit full patent to read further.

URL and more information on this patent, see: Gout, Michael P. Token based communications for machine learning systems. U.S. Patent Number 11847246, filed September 12, 2018, and published online on December 19, 2023. Patent URL (for desktop use only): https://ppubs.uspto.gov/pubwebapp/external.html?q=(11847246)&db=USPAT&type=ids

(Our reports deliver fact-based news of research and discoveries from around the world.)