Generating skincare product recommendations for a user based on skincare product attributes and user location and demographic data
AssigneeLife Spectacular, Inc., d/b/a Proven Skincare
InventorsZaoshi Amy Yuan, Ming S. Zhao
Life of this patent
Estimating the term…
Abstract
Systems and methods for generating user-specific recommendations of products or services by collaborative filtering executed and/or performed by one or more trained or untrained predictive models configured to ingest product attribute(s), product purpose(s), user location data, and/or user demographics. The predictive model(s) are leveraged to detect and determine user-specific preferences for, and preferences against, particular attributes, features, ingredients, aesthetic styles, and so on.
Claims20
A method of operating a host server coupled to a client device operated by a user to diagnose one or more skin concerns of the user and to identify a treatment for the one or more skin concerns, the method comprising:
- receiving from the client device at the host server, a user dataset comprising an input provided by the user to the client device;
- for each respective skin concern identifier of a set of skin concern identifiers accessible to a database service of the host server:
- query the database service to retrieve a set of diagnostics comprising data representing one or more signs and/or symptoms indicating a skin concern identified by the respective skin concern identifier;
- determine, for each respective diagnostic of the set of diagnostics, a credibility metric representing a likelihood based on the user dataset that the user exhibits the respective diagnostic;
- and
- determine, based on each credibility metric, a confidence metric representing a likelihood that the user does exhibit the respective skin concern;
- querying the database service to retrieve a skin concern co-occurrence dataset;
- ingesting by a predictive model service of the host server, the user dataset, each confidence metric, and the skin concern co-occurrence dataset;
- receiving from the predictive model service, a diagnosis comprising a set of co-occurring skin concerns exhibited by the user;
- for each respective skin concern of the set of co-occurring skin concerns, query the database service to retrieve a respective list of active ingredients that each exhibits a therapeutic effect for users exhibiting the respective skin concern;
- and
- communicating a treatment recommendation to the client device, the treatment recommendation based on each list of active ingredients.
A method of operating a host server coupled to a client device to identify a skincare product for treatment for a skin concern, the method comprising:
- obtaining, by the host server, a user dataset comprising:
- demographic information of a user operating the client device;
- and
- answers provided by the user to a diagnostic survey;
- ingesting the user dataset by a diagnostic prediction model hosted by the host server;
- receiving from the diagnostic prediction model a diagnostic matrix, each entry corresponding to a predicted likelihood that the user exhibits a specific skin concern;
- ingesting the user dataset and the diagnostic matrix by a customer review prediction model hosted by the host server;
- receiving from the customer review prediction model an ingredient sentiment prediction matrix, each entry corresponding to a predicted likelihood that the user will submit a positive-sentiment product review of a skincare product containing a specified active ingredient;
- selecting, from the ingredient sentiment prediction matrix, a set of entries exceeding a threshold predicted likelihood of a positive-sentiment product review;
- defining a user-specific active ingredient list containing at least a threshold number of specified active ingredients associated with the set of entries selected from the ingredient sentiment prediction matrix;
- customizing a skincare product based on the user-specific active ingredient;
- and
- instructing, by the host server, the client device to display a suggestion to the user to purchase and use the customized skincare product.
A method of operating a host server to generate training data to train a customer review prediction model configured to output a predicted likelihood of a positive-sentiment review of a consumer product when the consumer product is used as treatment for a specified medical condition, the method comprising:
- obtaining, by the host server, web page text content served by a web server in communication with the host server, the web page text content containing a customer review;
- extracting, by the host server, text associated with the customer review from the web page text content;
- identifying, by the host server, the consumer product referenced by the customer review from the extracted text;
- identifying, by the host server, a medical condition referenced by the customer review from the extracted text;
- identifying, by the host server, an author of the customer review from the extracted text;
- extracting, by the host server, a scaled rating of the consumer product by the author of the customer review from the extracted text;
- determining, by the host server, demographic information of the author of the customer review from the extracted text;
- querying, by the host server, a product database to obtain an ingredient list of the consumer product;
- inserting, by the host server, into a predictive model training data database in communication with the host server:
- the consumer product;
- the specified medical condition;
- the author of the customer review;
- the demographic information of the author;
- the scaled rating of the consumer product;
- and
- the ingredient list;
- and
- initiating a process to train a predictive model from the predictive model training data database.
Dependencies are inferred from the claim text.