A system and method for machine learning-based embedding and clustering of documents with subsequent action ranking and notification.

Abstract

This provides exemplary methods, systems, and storage media for using machine learning components that possess knowledge of multiple numerical embeddings corresponding to multiple documents. [Solution] The method includes obtaining constraints indicating the type of period and subject, and using the constraints to obtain a set of documents and corresponding metadata; for the set of documents, generating a set of numerical embeddings; generating an updated set of similarity clusters by updating the set of similarity clusters to incorporate the set of numerical embeddings; identifying a set of potential actions based on links between entity information and the updated set of similarity clusters; ranking the set of potential actions; and providing notifications indicating the higher-ranked actions.

Claims

1. A method performed with machine learning components including one or more machine learning models, The machine learning component obtains constraints indicating the period and subject type, wherein the machine learning component has knowledge of multiple numerical embeddings corresponding to multiple documents, and the multiple numerical embeddings are obtained relating to a set of similarity clusters. Using the aforementioned constraints, retrieve a set of documents and their corresponding metadata, wherein at least a subset of the set of documents is not included in the plurality of documents. Based on the set of documents and the information contained in the corresponding metadata, a set of numerical embeddings for the set of documents is generated. The process involves updating the set of similarity clusters to incorporate the aforementioned set of numerical embeddings, thereby generating an updated set of similarity clusters. For an entity, retrieve entity information including one or more sets of workflows and one or more datasets, Based on the links between the entity information and the updated set of similarity clusters, identify a set of potential actions. A set of ranked potential actions is generated by ranking the set of potential actions according to one or more criteria. A method comprising providing the entity with a notification indicating a set of higher-ranked potential actions from the aforementioned ranked set of potential actions.

2. The method according to claim 1, wherein the corresponding metadata is a subset of metadata available for the set of documents, and the corresponding metadata is identified from the available metadata using the constraints.

3. The method according to claim 1, wherein the set of documents comprises a plurality of data types, and the machine learning component is configured to capture a plurality of modalities of input data.

4. The method according to claim 1, further comprising applying a clustering algorithm to the plurality of numerical embeddings to form the set of similarity clusters.

5. The method according to claim 1, wherein the updated set of similarity clusters is formed by applying a clustering algorithm to the plurality of numerical embeddings and the set of numerical embeddings.

6. The method according to claim 1, further comprising ranking the updated set of similarity clusters based on one or more interest metrics.

7. The method according to claim 1, wherein the entity information indicates the subjective capabilities of the entity.

8. The method according to claim 1, further comprising identifying a set of entities linked to the set of higher-ranked potential actions, wherein the set of entities is identified based on the respective entity data of each entity in the set of entities.

9. The method according to claim 1, further comprising correlating the updated set of similarity clusters with clinical trial data, wherein the set of potential actions is identified based on the clinical trial data.

10. The method according to claim 1, further comprising correlating the updated set of similarity clusters with patient cohort data, wherein the set of potential actions is identified based on the patient cohort data.

11. The method according to claim 1, wherein the constraint is received as part of a user request or as part of an automated workflow.

12. The one or more criteria for ranking the set of potential actions are, Criteria related to the evaluation value for completing the corresponding potential action, Criteria related to the assessed ability of the entity to carry out the corresponding potential actions, and The method according to claim 1, comprising one or more criteria relating to the estimated cost of the entity to perform the corresponding potential action.

13. After generating the updated set of similarity clusters, Receiving user queries requesting information about the updated set of similarity clusters, The method according to claim 1, further comprising providing information from the set of documents in response to the user query.

14. A computing system, One or more processors, Memory and, A computing system comprising one or more programs stored in the memory and configured to be executed by one or more processors, wherein one or more programs include instructions that cause or trigger the execution of the method described in any one of claims 1 to 13.

15. A computer-readable storage medium comprising executable instructions, when executed by one or more processors, that cause the one or more processors to perform or trigger the performance of the method described in any one of claims 1 to 13.

Images