Granted Patent B2

US 12,273,304 · App. 17/244,846 · Granted Apr 8, 2025

Inventors: Andrew R. Freed (Cary, NC); Lalit Agarwalla (Bangalore, IN)

Assignee: International Business Machines Corporation

H04L51/02G06N5/01

Abstract

Claims (47)

1. A method for altering an automated conversation system, the method comprising:

receiving, by one or more computer processors, automatic conversation log data;

generating, by the one or more computer processors, a decision tree from the log data;

summarizing, by the one or more computer processors, child node data for a node of the decision tree, wherein summarizing the child node data for the node comprises summarizing a distribution of parent node intents, entities, and context variable values, and child node response keywords; and

generating, by the one or more computer processors, node response variations according to the child node data of the node by:

generating alternative flows by changing an element selected from a group consisting of Boolean logic, operators, and combinations thereof, of conditional expressions of the node;

determining changes associated with conditional results percentages for each generated alternative flow; and

retaining only alternative flows having changes in conditional results percentages exceeding a defined threshold.

2. The method according to claim 1 , wherein the log data comprises user conversation turns, user utterances, chatbot responses, and previous nodes visited.

3. The method according to claim 1 , further comprising:

annotating, by the one or more computer processors, conversation turns of the node; and

generating, by the one or more computer processors, conversation turn variations according to the annotations of the conversation turns of the node.

4. The method according to claim 3 , wherein annotating the conversation turns for the node comprises annotating a number of visits to the node, and the decision for each visit.

5. The method according to claim 3 , wherein generating the conversation turn variations comprises generating variations according to node decision results.

6. The method according to claim 1 , wherein generating node response variations comprises generating responses according to intents and/or entities.

7. A computer program product for altering an automated conversation system, the computer program product comprising one or more computer readable storage devices and collectively stored program instructions on the one or more computer readable storage devices, the stored program instructions comprising:

program instructions to receive automatic conversation log data;

program instructions to generate a decision tree from the data;

program instructions to summarize child node data for a node of the decision tree, wherein summarizing the child node data for the node comprises summarizing a distribution of parent node intents, entities, and context variable values, and child node response keywords; and

program instructions to generate node response variations according to the child node data of the node by:

generating alternative flows by changing an element selected from a group consisting of Boolean logic, operators, and combinations thereof, of conditional expressions of the node;

determining changes associated with conditional results percentages for each generated alternative flow; and

retaining only alternative flows having changes in conditional results percentages exceeding a defined threshold.

8. The computer program product according to claim 7 , wherein the conversation log data comprises user conversation turns, user utterances, chatbot responses, and previous nodes visited.

9. The computer program product according to claim 7 , the stored program instructions further comprising:

program instructions to annotate conversation turns of the node; and

program instructions to generate conversation turn variations according to the annotations of the conversation turns of the node.

10. The computer program product according to claim 9 , wherein annotating the conversation turns for the node comprises annotating a number of visits to the node, and the decision for each visit.

11. The computer program product according to claim 9 , wherein generating the conversation turn variations comprises generating variations according to node decision results.

12. The computer program product according to claim 7 , wherein generating node response variations comprises generating responses according to intents and/or entities.

13. A computer system for altering an automated conversation system, the computer system comprising:

one or more computer processors;

one or more computer readable storage devices; and

stored program instructions on the one or more computer readable storage devices for execution by the one or more computer processors, the stored program instructions comprising:

program instructions to receive automatic conversation log data;

program instructions to generate a decision tree from the data;

program instructions to summarize child node data for a node of the decision tree, wherein summarizing the child node data for the node comprises summarizing a distribution of parent node intents, entities, and context variable values, and child node response keywords; and

program instructions to generate node response variations according to the child node data of the node by:

generating alternative flows by changing an element selected from a group consisting of Boolean logic, operators, and combinations thereof, of conditional expressions of the node;

determining changes associated with conditional results percentages for each generated alternative flow; and

retaining only alternative flows having changes in conditional results percentages exceeding a defined threshold.

14. The computer system according to claim 13 , wherein the conversation log data comprises user conversation turns, user utterances, chatbot responses, and previous nodes visited.

15. The computer system according to claim 13 , the stored program instructions further comprising:

program instructions to annotate conversation turns of the node; and

program instructions to generate conversation turn variations according to the annotations of the conversation turns of the node.

16. The computer system according to claim 15 , wherein annotating the conversation turns for the node comprises annotating a number of visits to the node, and the decision for each visit.

17. The computer system according to claim 15 , wherein generating the conversation turn variations comprises generating variations according to node decision results.

Assignments (1)

Continuity (1)

References Cited (58)