Granted Patent B2

US 12,555,689 · App. 18/452,417 · Granted Feb 17, 2026

Inventors: Natalia Mulligan (Dublin, IE); Marco Luca Sbodio (Dublin, IE); Joao H. Bettencourt-Silva (Dublin, IE); Vanessa Lopez Garcia (Dublin, IE); Gabriele Picco (Dublin, IE); Marcos Martínez Galindo (Dublin, IE)

Assignee: International Business Machines Corporation

G16H50/70G06Q10/06315G16H10/20

Abstract

Claims (65)

1 . A computer-implemented method for generating and ranking potential barriers that prevent completion of a structured process, comprising:

receiving a textual description of the structured process;

receiving one or more templates relating to the structured process, wherein the template comprises an explanation for why a user action required in the process could not be completed;

identifying a set of user actions required to complete the structured process based on the textual description;

expanding the set of user actions based on general knowledge to include one or more additional actions;

updating the one or more templates based on the expanded set of user actions;

generating, using at least one generative language model, a set of potential barriers that prevent completion of the structured process based on the one or more updated templates;

fine-tuning the at least one generative language model on domain textual data and feedback on barriers in the set of potential barriers indicating that the barriers were incorrectly detected; and

generating, using the fine-tuned generative language model, a second set of potential barriers.

2 . The method of claim 1 , further comprising:

determining, for each potential barrier, a corresponding confidence score using a learned probability distribution of the at least one generative language model;

ranking the set of potential barriers based on each confidence score corresponding to each potential barrier; and

providing a ranked list based on the ranking as output.

3 . The method of claim 2 , further comprising:

receiving user feedback on the ranked list; and

fine-tuning the at least one generative language model based on the user feedback.

4 . The method of claim 1 , wherein the textual description describes the set of user actions that one or more participants enrolled or participating in the structured process are required to perform to complete the structured process.

5 . The method of claim 1 , wherein identifying the set of user actions based on the textual description comprises:

applying natural language processing (NLP) to the textual description.

6 . The method of claim 1 , wherein the general knowledge comprises at least one of knowledge graphs and deep learning question answering and information retrieval models.

7 . The method of claim 1 , wherein the one or more additional user actions are required to complete the set of user actions.

8 . The method of claim 1 , wherein updating the one or more templates based on the expanded set of user actions comprises:

filling in one or more fields of the one or more templates based on the expanded set of user actions.

9 . A system for generating and ranking potential barriers that prevent completion of a structured process, comprising:

at least one processor; and

a processor-readable memory device storing instructions that when executed by the at least one processor causes the at least one processor to perform operations including:

receiving a textual description of the structured process;

receiving one or more templates relating to the structured process, wherein the template comprises an explanation for why a user action required in the process could not be completed;

identifying a set of user actions required to complete the structured process based on the textual description;

expanding the set of user actions based on general knowledge to include one or more additional actions;

updating the one or more templates based on the expanded set of user actions;

generating, using at least one generative language model, a set of potential barriers that prevent completion of the structured process based on the one or more updated templates;

fine-tuning the at least one generative language model on domain textual data and feedback on barriers in the set of potential barriers indicating that the barriers were incorrectly detected; and

generating, using the fine-tuned generative language model, a second set of potential barriers.

10 . The system of claim 9 , wherein the operations further include:

determining, for each potential barrier, a corresponding confidence score using a learned probability distribution of the at least one generative language model;

ranking the set of potential barriers based on each confidence score corresponding to each potential barrier; and

providing a ranked list based on the ranking as output.

11 . The system of claim 10 , wherein the operations further include:

receiving user feedback on the ranked list; and

fine-tuning the at least one generative language model based on the user feedback.

12 . The system of claim 9 , wherein the textual description describes the set of user actions that one or more participants enrolled or participating in the structured process are required to perform to complete the structured process.

13 . The system of claim 9 , wherein identifying the set of user actions based on the textual description comprises:

applying natural language processing (NLP) to the textual description.

14 . The system of claim 9 , wherein the general knowledge comprises at least one of knowledge graphs and deep learning question answering and information retrieval models.

15 . The system of claim 9 , wherein the one or more additional user actions are required to complete the set of user actions.

16 . The system of claim 9 , wherein updating the one or more templates based on the expanded set of user actions comprises:

filling in one or more fields of the one or more templates based on the expanded set of user actions.

17 . A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:

receive a textual description of a structured process;

receive one or more templates relating to the structured process, wherein the template comprises an explanation for why a user action required in the process could not be completed;

identify a set of user actions required to complete the structured process based on the textual description;

expand the set of user actions based on general knowledge to include one or more additional actions;

update the one or more templates based on the expanded set of user actions;

generate, using at least one generative language model, a set of potential barriers that prevent completion of the structured process based on the one or more updated templates;

fine-tuning the at least one generative language model on domain textual data and feedback on barriers in the set of potential barriers indicating that the barriers were incorrectly detected; and

generate, using the fine-tuned generative language model, a second set of potential barriers.

18 . The computer program product of claim 17 , wherein the program instructions executable by the processor further cause the processor to:

determine, for each potential barrier, a corresponding confidence score using a learned probability distribution of the at least one generative language model;

rank the set of potential barriers based on each confidence score corresponding to each potential barrier; and

provide a ranked list based on the ranking as output.

19 . The computer program product of claim 18 , wherein the program instructions executable by the processor further cause the processor to:

receive user feedback on the ranked list; and

fine-tune the at least one generative language model based on the user feedback.

20 . The computer program product of claim 17 , wherein the textual description describes the set of user actions that one or more participants enrolled or participating in the structured process are required to perform to complete the structured process.

Assignments (1)

Continuity (1)

References Cited (35)