Approaches to hybrid learning of personalized recommendations and interactive digital health system for implementing the same
Inventors
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12450230-B1
Publication Date
2025-10-21
Expiration Date
Abstract
Introduced here is an interactive digital health system (or simply “interactive system”) that is able to update a knowledge base used to facilitate semi-automated conversations to account for feedback. The interactive system can include a collection of modules that, in operation, can provide or support coaching services in the digital health domain, Upon receiving input indicative of a question from a patient, these modules allow the interactive system to perform language modeling to identify content objects in a knowledge base that are relevant to the question. Moreover, the interactive system may update the knowledge base to account for feedback, thereby making the knowledge base more robust to a broad range of questions.
Core Innovation
The invention relates to an interactive digital health coaching system that uses transformer model sentence embeddings to semantically match a query from a first individual to health-related media content stored as content objects in a knowledge base. A query is provided to a transformer model to produce a query-associated embedding representing an enumeration of the query in a language domain, and the query-associated embedding is compared against embeddings associated with content objects in the knowledge base.
Based on the comparing, a second set of embeddings that most closely corresponds to the query-associated embedding is identified and the associated content objects are presented on a device-implemented coach console accessible to a second individual. The coach console acquires input indicative of a selection of a content object, and the selected content object’s content is used to add a new content object programmatically linked to the query-associated embedding in the knowledge base.
In the broader described approach, the knowledge base includes entries that provide a recommendation to be implemented for improving health, an educational summary of a disease or treatment, and/or a motivational statement. The selected content is added so that the new content object can be automatically suggested for a future query that is identical or similar to the query.
Claims Coverage
Independent claim set covers processor-implemented query-to-embedding matching, selection by a second individual, and feedback-driven knowledge-base expansion so content is automatically suggested for identical or similar future queries. Across the independent claims, four main inventive capability areas are recited.
Transformer-model query embedding matching to knowledge-base content objects
providing the query to a transformer model to produce a query-associated embedding, comparing the query-associated embedding against embeddings associated with content objects included in a knowledge base used for responding to queries, and identifying based on the comparing a second set of embeddings that most closely correspond to the query-associated embedding
Coach-console presentation and selection-driven knowledge-base expansion linked to the query embedding
causing digital presentation of content objects associated with the second set of embeddings on a device-implemented coach console accessible to a second individual, acquiring input indicative of a selection of a content object, and adding a new content object programmatically linked to the query-associated embedding so the new content object can be automatically suggested for a future query identical or similar to the query
Semi-automated knowledge-base response using multiple content-bearing data structures linked to query embeddings
identifying multiple data structures from among a set of data structures maintained in a knowledge base used for responding to queries in a semi-automated manner, acquiring input indicative of a selection of a data structure, and adding a new data structure programmatically linked to the embedding produced for the query
Personalizing knowledge-base conversation content with recommendation, educational summary, and motivational statement entries linked to query embeddings
comparing an embedding produced for the query against embeddings produced for entries in the knowledge base to identify multiple embeddings that most closely correspond to the query, where each entry includes a recommendation to be implemented for improving health, an educational summary of a disease or a treatment, or a motivational statement, and adding a new entry programmatically linked to the embedding produced for the query
Across the independent claims, the invention centers on generating query-associated embeddings using a transformer model, comparing the query embedding to embeddings for knowledge-base content objects, entries, or data structures containing health-related media content, presenting multiple closely corresponding items to a second individual for selection, and expanding the knowledge base by adding new content programmatically linked to the query embedding so that selected content can be automatically suggested for future identical or similar queries.
Stated Advantages
Automatically suggests a new content object for a future query that is identical or similar to the query.
Facilitates conversations by personalizing a knowledge base to include selected content tied to embeddings for future identical or similar queries.
Enables presenting multiple closely corresponding knowledge-base entries associated with embeddings that most closely correspond to the query.
Documented Applications
Interactive digital health coaching where a coach console presents health-related media content cards associated with closely matching query embeddings for selection by a second individual.
Facilitating conversations using a personalized knowledge base in which entries include a recommendation, educational summary, and/or motivational statement for improving health of the first individual, with selected content used to expand the knowledge base for future query suggestions.
Interested in licensing this patent?