Systems and methods for regulating provision of messages with content from disparate sources based on risk and feedback data
Inventors
Guttikonda, Sudheer • Morse, William • Qiu, Chuanhan • Speier, Austin
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12315611-B2
Publication Date
2025-05-27
Expiration Date
Abstract
Aspects of the present disclosure are directed to systems, methods, and computer readable media for regulating digital therapeutic content for provision. A computing system may identify a digital therapeutic content to be provided via a network. The computing system may apply the digital therapeutic content to a machine learning (ML) model having a set of weights. The computing system may determine, from applying the digital therapeutic content to the ML model, an indication as of one of compliance or non-compliance. The computing system may store, using one or more data structures, an association between the digital therapeutic content and the indication used to control provision of the digital therapeutic content via the network, by (i) restricting the digital therapeutic content from provision responsive to determining the indication of non-compliance and (ii) permitting the digital therapeutic content to be provided responsive to determining the indication of compliance.
Core Innovation
The invention relates to identifying first digital content to be provided via a network and applying the first digital content to a machine learning (ML) model having a set of weights to generate a first output. The ML model is trained by identifying a training dataset including a plurality of examples, where each example identifies respective second digital content and a first indication identifying one of compliance or non-compliance for provision, comparing the first indication with a second indication determined from applying the second digital content to the ML model, and updating at least one of the set of weights responsive to the comparing.
After applying the ML model to the first digital content and determining from the first output an indication of the first digital content as non-compliant, the invention stores using one or more data structures an association between the first digital content and the indication. This association restricts the first digital content from provision responsive to determining the indication as non-compliant, and the invention generates a portion identifying at least a subsection of the first digital content to be modified responsive to determining the indication as non-compliant.
In additional aspects, the invention includes restricting first digital content from provision when non-compliant and permitting the first digital content to be provided when compliant, while determining a second indication from the output used to control provision. The invention further supports determining domain-based compliance or non-compliance by selecting an ML model corresponding to at least one domain, and receiving through a user interface a user-provided third indication that overrides the indication output by the ML model.
Claims Coverage
The independent claims cover training an ML model for compliance/non-compliance determinations, using the resulting indications to control provision, storing restrictions as associations, and generating a portion identifying a subsection for modification. Across the independent claims, the core inventive features include model training from labeled examples with updating based on comparisons, generating non-compliance indications for restriction, and generating modification portions; additional inventive features include permitting versus restricting based on compliance, user-interface override, and domain-based selection among multiple ML models.
Compliance label-based ML model training with weight updating
Identifying a training dataset including a plurality of examples, each example identifying respective second digital content and a first indication identifying one of compliance or non-compliance for provision, applying the second digital content into the ML model to generate a second output, determining a second indication of the second digital content as one of compliant or non-compliant used to control provision, comparing the first indication with the second indication, and updating responsive to comparing the first indication with the second indication, at least one of the set of weights of the ML model using the comparison to further train the ML model.
Non-compliance indication to restrict provision and generate modification portion
Determining, from the first output, an indication of the first digital content as non-compliant; storing an association between the first digital content and the indication to restrict the first digital content from provision responsive to determining the indication of the first digital content as non-compliant; and generating, based on applying the ML model, a portion identifying at least a subsection of the first digital content to be modified, responsive to determining the indication of the first digital content as non-compliant.
Compliance-based control of provision with ML-indication comparison
Restricting the first digital content from provision when non-compliant and permitting the first digital content to be provided when compliant; determining, from the output, a second indication of the first digital content as one of compliant or non-compliant used to control provision; comparing the first indication from the example of the plurality of examples of the training dataset with the second indication determined by the ML model; and updating responsive to comparing the first indication with the second indication, at least one of the set of weights of the ML model using the comparison to further update the ML model.
User interface override of compliance/non-compliance indication
Receiving, through a user interface, a user-provided third indication that marks the first digital content as compliant or non-compliant, and overriding the indication output by the machine learning (ML) model with that third indication.
Domain-based selection and domain-specific non-compliance determination
Receiving an identification of first digital content associated with at least one domain, selecting an ML model corresponding to the at least one domain from a plurality of ML models, and determining an indication that the first digital content is non-compliant with respect to the at least one domain.
Across the independent claims, the patent claims an end-to-end process where labeled compliance/non-compliance examples train an ML model via comparison-based weight updates, the ML model outputs compliance/non-compliance indications used to restrict or permit provision, the system stores associations to reflect restrictions, and the system generates a portion identifying a subsection to be modified when non-compliance is determined. Additional independent-claim refinements include compliance-based permitting/restricting, user-interface override of the ML decision, and domain-based selection among multiple ML models.
Stated Advantages
Restricting digital content from provision responsive to determining the indication as non-compliant.
Permitting the digital content to be provided when compliant.
Generating a portion identifying at least a subsection to be modified responsive to determining non-compliance.
Further training the ML model responsive to comparing a training dataset indication with the ML model-determined indication.
Determining compliance or non-compliance based on at least one domain by selecting an ML model corresponding to the domain.
Overriding the indication output by the ML model with a user-provided third indication received via a user interface.
Documented Applications
No documented applications found
Interested in licensing this patent?