Abstract: A health provider matching service is provided to match patients to health providers based on a semantic relationship graph of data associated with conditions of patients. Using natural language processing, the service identifies terms describing symptoms, treatments, and health providers associated with a condition. The service then identifies semantic relations among the terms and probability distributions of the terms to generate a semantic relationship graph. Quality information of therapists is combined with machine learning techniques to identify features that are used to distinguish high quality and low quality providers. Based on the semantic relationship graph and the derived quality information, the service matches a patient to a health provider suitable to treat the particular needs of the patient. The service also handles social, behavioral, or emotional issues for which medications are not appropriate.

Abstract: An optimization system optimizes groups of providers for a given population of users. The optimization system includes various engines such as a source interface engine, selection engine, and group score engine. The source interface engine receives information to be used in constrained optimization from server computers and stores the information in databases. The selection engine selects groups of providers from a set of all available providers. The group score engine generates a score for each group of providers. The selection engine continues to iteratively select groups of providers to minimize the generated score. The selection engine can use hard constraints, for example, requiring that a certain type of provider be included in all selected groups, or soft constraints, for example, requiring that a certain number of providers be included based on a size of the given population.

Abstract: An optimization system optimizes groups of providers for a given population of users. The optimization system includes various engines such as a source interface engine, selection engine, and group score engine. The source interface engine receives information to be used in constrained optimization from server computers and stores the information in databases. The selection engine selects groups of providers from a set of all available providers. The group score engine generates a score for each group of providers. The selection engine continues to iteratively select groups of providers to minimize the generated score. The selection engine can use hard constraints, for example, requiring that a certain type of provider be included in all selected groups, or soft constraints, for example, requiring that a certain number of providers be included based on a size of the given population.

Abstract: A health provider matching service is provided to match patients to health providers based on a semantic relationship graph of data associated with conditions of patients. Using natural language processing, the service identifies terms describing symptoms, treatments, and health providers associated with a condition. The service then identifies semantic relations among the terms and probability distributions of the terms to generate a semantic relationship graph. Quality information of therapists is combined with machine learning techniques to identify features that are used to distinguish high quality and low quality providers. Based on the semantic relationship graph and the derived quality information, the service matches a patient to a health provider suitable to treat the particular needs of the patient. The service also handles social, behavioral, or emotional issues for which medications are not appropriate.