Domain-specific generative machine learning models

1. A computer program product, the computer program product being tangibly embodied on a non-transitory computer-readable storage medium and comprising instructions that, when executed by at least one computing device, are configured to cause the at least one computing device to:

receive a plurality of resolved incident tickets of an incident domain, each resolved incident ticket having a worklog providing a history of actions taken during attempts to resolve a corresponding resolved incident and a resolution having at least one resolution statement for the corresponding resolved incident;

execute an iterative processing of the plurality of resolved incident tickets to obtain processed incident tickets, including:

(a) processing each resolution statement of the resolution with at least one domain-specific statement classifier that is specific to the incident domain to either discard or retain a classified resolution statement,

(b) processing each retained classified resolution statement in conjunction with the worklog to determine whether to discard or retain the resolved incident,

(c) providing an updated resolution for the resolved incident when the resolved incident is retained, and

(d) adding the resolved incident with the updated resolution to the processed incident tickets; and

train at least one machine learning model to process a new incident ticket including generating a new resolution statement using a new worklog of the new incident ticket, using the processed incident tickets.

2. The computer program product of claim 1 , wherein the at least one machine learning model includes a domain-specific large language model (LLM).

3. The computer program product of claim 1 , wherein the at least one domain-specific statement classifier includes a noisy statement classifier configured to classify each resolution statement as noise with respect to the incident domain.

4. The computer program product of claim 1 , wherein the at least one domain-specific statement classifier includes a knowledge base classifier configured to classify each resolution statement as referencing a knowledge base for the incident domain.

5. The computer program product of claim 1 , wherein the at least one domain-specific statement classifier includes a domain ontology classifier configured to classify each resolution statement as including domain vocabulary for the incident domain.

6. The computer program product of claim 1 , wherein the instructions, when executed, are further configured to cause the at least one computing device to:

determine whether to discard or retain the resolved incident by inspecting each retained classified resolution statement for inclusion of a resolution wake word.

7. The computer program product of claim 1 , wherein the instructions, when executed, are further configured to cause the at least one computing device to:

determine whether to discard or retain the resolved incident by inspecting the worklog for inclusion of a resolution wake word.

8. The computer program product of claim 1 , wherein the instructions, when executed, are further configured to cause the at least one computing device to:

determine whether to discard or retain the resolved incident by determining a similarity level between the retained classified resolution statement and at least one worklog statement of the worklog.

9. The computer program product of claim 1 , wherein the instructions, when executed, are further configured to cause the at least one computing device to:

determine whether to discard or retain the resolved incident based on an incident agent quality rating of an incident agent associated with the resolved incident ticket.

10. The computer program product of claim 1 , wherein the instructions, when executed, are further configured to cause the at least one computing device to:

determine that a quantity of retained incident tickets with updated resolutions is below a threshold;

synthetically generate at least one additional resolved incident ticket using at least one of a discarded incident ticket or a retained incident ticket; and

include the at least one additional resolved incident ticket in the processed incident tickets.

11. A computer-implemented method, the method comprising:

receiving a plurality of resolved incident tickets of an incident domain, each resolved incident ticket having a worklog providing a history of actions taken during attempts to resolve a corresponding resolved incident and a resolution having at least one resolution statement for the corresponding resolved incident;

executing an iterative processing of the plurality of resolved incident tickets to obtain processed incident tickets, including:

(a) processing each resolution statement of the resolution with at least one domain-specific statement classifier that is specific to the incident domain to either discard or retain a classified resolution statement,

(b) processing each retained classified resolution statement in conjunction with the worklog to determine whether to discard or retain the resolved incident,

(c) providing an updated resolution for the resolved incident when the resolved incident is retained, and

(d) adding the resolved incident with the updated resolution to the processed incident tickets; and

training at least one machine learning model to process a new incident ticket including generating a new resolution statement using a new worklog of the new incident ticket, using the processed incident tickets.

12. The method of claim 11 , wherein the at least one machine learning model includes a domain-specific large language model (LLM).

13. The method of claim 11 , wherein the at least one domain-specific statement classifier includes at least one of: a noisy statement classifier configured to classify each resolution statement as noise with respect to the incident domain, a knowledge base classifier configured to classify each resolution statement as referencing a knowledge base for the incident domain, and a domain ontology classifier configured to classify each resolution statement as including domain vocabulary for the incident domain.

14. The method of claim 11 , further comprising:

determining whether to discard or retain the resolved incident by inspecting each retained classified resolution statement and the worklog for inclusion of a resolution wake word.

15. The method of claim 11 , further comprising:

determining whether to discard or retain the resolved incident by determining a similarity level between the retained classified resolution statement and at least one worklog statement of the worklog.

16. The method of claim 11 , further comprising:

determining that a quantity of retained incident tickets with updated resolutions is below a threshold;

synthetically generating at least one additional resolved incident ticket using at least one of a discarded incident ticket or a retained incident ticket; and

including the at least one additional resolved incident ticket in the processed incident tickets.

17. A system comprising:

at least one memory including instructions; and

at least one processor that is operably coupled to the at least one memory and that is arranged and configured to execute instructions that, when executed, cause the at least one processor to:

receive a plurality of resolved incident tickets of an incident domain, each resolved incident ticket having a worklog providing a history of actions taken during attempts to resolve a corresponding resolved incident and a resolution having at least one resolution statement for the corresponding resolved incident;

execute an iterative processing of the plurality of resolved incident tickets to obtain processed incident tickets, including:

(a) processing each resolution statement of the resolution with at least one domain-specific statement classifier that is specific to the incident domain to either discard or retain a classified resolution statement,

(b) processing each retained classified resolution statement in conjunction with the worklog to determine whether to discard or retain the resolved incident,

(c) providing an updated resolution for the resolved incident when the resolved incident is retained, and

(d) adding the resolved incident with the updated resolution to the processed incident tickets; and

train at least one machine learning model to process a new incident ticket including generating a new resolution statement using a new worklog of the new incident ticket, using the processed incident tickets.

18. The system of claim 17 , wherein the at least one domain-specific statement classifier includes at least one of: a noisy statement classifier configured to classify each resolution statement as noise with respect to the incident domain, a knowledge base classifier configured to classify each resolution statement as referencing a knowledge base for the incident domain, and a domain ontology classifier configured to classify each resolution statement as including domain vocabulary for the incident domain.

19. The system of claim 17 , wherein the instructions, when executed, are further configured to cause the at least one processor to:

determine whether to discard or retain the resolved incident by inspecting each retained classified resolution statement and the worklog for inclusion of a resolution wake word.

20. The system of claim 17 , wherein the instructions, when executed, are further configured to cause the at least one processor to:

determine that a quantity of retained incident tickets with updated resolutions is below a threshold;

synthetically generate at least one additional resolved incident ticket using at least one of a discarded incident ticket or a retained incident ticket; and

include the at least one additional resolved incident ticket in the processed incident tickets.