Customization of machine learning tools with occupation training
Methods and apparatus are disclosed for customizing a copilot or other machine learning tool. Document records related to copilot objectives or tasks are obtained and used to identify corresponding data sources. Data sources can be integrated into data producers or data repositories, to be used by a retrieval microservice. Data producers and other microservices are individually fine-tuned for the custom application, before or after integration into the copilot. The integrated copilot is tested end-to-end, and can be further refined. Disclosed techniques range from fully automated to human-in-the-loop (e.g. guided by an expert) to fully interactive. Some techniques produce tools which can automate certain customization operations. Training data, tasks, or document records blend expert-generated, developer-generated, or synthesized items.
1. A method comprising:
(a) producing a language-trained machine learning tool ML 2 from a machine learning tool ML 1 by training the machine learning tool ML 1 to perform general language tasks with at least a first predetermined performance level;
(b) producing an occupation-trained machine learning tool ML 3 from the language-trained machine learning tool ML 2 by training the machine learning tool ML 2 to perform occupation-related tasks of a given occupation with at least a second predetermined performance level; and
(c) producing a custom-trained machine learning tool ML 4 from the occupation-trained machine learning tool ML 3 by training the machine learning tool ML 3 to perform custom tasks of the given occupation with at least a third predetermined performance level.
2. The method of claim 1 , wherein the custom tasks of the given occupation or the occupation-related tasks comprise identifying one or more pertinent document records from a data corpus in response to a given task description, and the method further comprises:
prompting the machine learning tool ML 3 or ML 4 with a task description within scope of a copilot objective; and
receiving the pertinent document record(s) from the machine learning tool ML 3 or ML 4 .
3. The method of claim 2 , further comprising:
validating a subset of the received document record(s).
4. The method of claim 1 , wherein the custom tasks of the given occupation or the occupation-related tasks comprise determining, in response to a given document record, one or more data sources which support the given document record, and the method further comprises:
prompting the machine learning tool ML 3 or ML 4 with a document record pertinent to a copilot objective; and
receiving the supporting data source(s) from the machine learning tool ML 3 or ML 4 .
5. The method of claim 4 , further comprising:
validating a subset of the received data source(s).
6. The method of claim 1 , wherein the given occupation comprises interviewing.
7. The method of claim 1 , wherein the given occupation comprises annotating interviews or annotating recorded work sessions.
8. The method of claim 1 , wherein the custom-trained machine learning tool ML 4 is a first custom-trained machine learning tool, and the method further comprises:
(d) producing a second custom-trained machine learning tool ML 5 from the occupation-trained machine learning tool ML 3 by training the machine learning tool ML 3 to perform second custom tasks of the given occupation with at least a fourth predetermined performance level.
9. The method of claim 8 , further comprising:
(e) deploying machine learning tools ML 4 and ML 5 at distinct first and second organizations respectively;
wherein the training at act (c) uses proprietary data of the first organization without using proprietary data of the second organization, and the training at act (d) uses proprietary data of the second organization without using proprietary data of the first organization.
10. The method of claim 1 , further comprising:
(d) deploying machine learning tool ML 4 within a first microservice of a copilot, the copilot comprising a weakly connected network of microservices including the first microservice.
11. One or more computer-readable media storing instructions which, when executed by one or more hardware processors, cause the one or more hardware processors to perform operations comprising:
(a) producing a language-trained machine learning tool ML 2 from a machine learning tool ML 1 by training the machine learning tool ML 1 to perform general language tasks with at least a first predetermined performance level;
(b) producing an occupation-trained machine learning tool ML 3 from the language-trained machine learning tool ML 2 by training the machine learning tool ML 2 to perform occupation-related tasks of a given occupation with at least a second predetermined performance level; and
(c) producing a custom-trained machine learning tool ML 4 from the occupation-trained machine learning tool ML 3 by training the machine learning tool ML 3 to perform custom tasks of the given occupation with at least a third predetermined performance level.
12. The one or more computer-readable media of claim 11 , wherein the custom tasks of the given occupation or the occupation-related tasks comprise identifying one or more pertinent document records from a data corpus in response to a given task description, and the operations further comprise:
prompting the machine learning tool ML 3 or ML 4 with a task description within scope of a copilot objective; and
receiving the pertinent document record(s) from the machine learning tool ML 3 or ML 4 .
13. The one or more computer-readable media of claim 12 , wherein the operations further comprise:
validating a subset of the received document record(s).
14. The one or more computer-readable media of claim 11 , wherein the custom-trained machine learning tool ML 4 is a first custom-trained machine learning tool, and the operations further comprise:
(d) producing a second custom-trained machine learning tool ML 5 from the occupation-trained machine learning tool ML 3 by training the machine learning tool ML 3 to perform second custom tasks of the given occupation with at least a fourth predetermined performance level.
15. A system, comprising:
one or more hardware processors with memory coupled thereto; and
computer-readable media storing instructions which, when executed by the one or more hardware processors, cause the one or more hardware processors to perform operations comprising:
(a) producing a language-trained machine learning tool ML 2 from a machine learning tool ML 1 by training the machine learning tool ML 1 to perform general language tasks with at least a first predetermined performance level;
(b) producing an occupation-trained machine learning tool ML 3 from the language-trained machine learning tool ML 2 by training the machine learning tool ML 2 to perform occupation-related tasks of a given occupation with at least a second predetermined performance level; and
(c) producing a custom-trained machine learning tool ML 4 from the occupation-trained machine learning tool ML 3 by training the machine learning tool ML 3 to perform custom tasks of the given occupation with at least a third predetermined performance level.
16. The system of claim 15 , wherein the custom tasks of the given occupation or the occupation-related tasks comprise determining, in response to a given document record, one or more data sources which support the given document record, and the operations further comprise:
prompting the machine learning tool ML 3 or ML 4 with a document record pertinent to a copilot objective; and
receiving the supporting data source(s) from the machine learning tool ML 3 or ML 4 .
17. The system of claim 16 , wherein the operations further comprise:
validating a subset of the received data source(s).
18. The system of claim 15 , wherein the custom-trained machine learning tool ML 4 is a first custom-trained machine learning tool, and the operations further comprise:
(d) producing a second custom-trained machine learning tool ML 5 from the occupation-trained machine learning tool ML 3 by training the machine learning tool ML 3 to perform second custom tasks of the given occupation with at least a fourth predetermined performance level.
19. The system of claim 15 , wherein the given occupation comprises interviewing.
20. The system of claim 15 , wherein the given occupation comprises annotating interviews or annotating recorded work sessions.