Graph based prediction for next action in conversation flow

Patent metadata

Field	Value
Publication number	US-11600276-B2
Application number	US-202117146256-A
Country	US
Kind code	B2
Filing date	Jan 11, 2021
Priority date	Sep 27, 2018
Publication date	Mar 7, 2023
Grant date	Mar 7, 2023

Abstract

One embodiment provides a method for predicting a next action in a conversation system that includes obtaining, by a processor, information from conversation logs and a conversation design. The processor further creates a dialog graph based on the conversation design. Weights and attributes for edges in the dialog graph are determined based on the information from the conversation logs and adding user input and external context information to an edge attributes set. An unrecognized user input is analyzed and a next action is predicted based on dialog nodes in the dialog graph and historical paths. A guiding conversation response is generated based on the predicted next action.

First claim

What is claimed is:

1. A method for predicting a next action in a conversation system comprising:
obtaining, by a processor, information from conversation logs and a conversation design;
creating, by the processor, a dialog graph based on the conversation design;
determining weights for edges in the dialog graph based on the information from the conversation logs and adding user input and external context information to an edge attributes set;
analyzing an unrecognized user input and predicting a next action based on dialog nodes in the dialog graph and historical paths;
generating a guiding conversation response based on the predicted next action;
improving the conversation design upon non-acceptance of the guiding conversation response based on analyzing the historical records including exception records, issue records and prediction results; and
updating dialog nodes and conversation transition logics based on historical records.

2. The method of claim 1 , further comprising:
updating the dialog graph based on the updated dialog nodes and transition logics.

3. The method of claim 2 , wherein:
the conversation design comprises all dialog node attributes and transition logics;
a node in the dialog graph comprises the dialog node and all its dialog node attributes; and
an edge in the dialog graph comprises transition logic for any given dialog node pairs.

4. The method of claim 1 , wherein the guiding conversation response is based on insertion of a temporary dialog node in the dialog graph.

5. The method of claim 4 , wherein determining weights for the edges comprises:
analyzing the conversation logs;
calculating transition probabilities among dialog nodes of the dialog graph;
weighting the edges based on the transition probabilities.

6. The method of claim 1 , wherein predicting the next action comprises determining whether a match exists for a particular dialog node in the dialog graph.

7. The method of claim 1 , wherein predicting the next action further comprises:
analyzing individual context information comprising: time, historical web page or mobile page visiting information, and location.

8. The method of claim 7 , wherein the individual context information is used for determining potential intents and for predicting the corresponding action.

9. A computer program product for predicting a next action in a conversation system, the computer program product comprising a non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:
obtain, by the processor, information from conversation logs and a conversation design;
create, by the processor, a dialog graph based on the conversation design;
determine, by the processor, weights for edges in the dialog graph based on the information from the conversation logs and adding user input and external context information to an edge attributes set;
analyze, by the processor, an unrecognized user input and predicting a next action based on dialog nodes in the dialog graph and historical paths;
generate, by the processor, a guiding conversation response based on the predicted next action;
improve, by the processor, the conversation design upon non-acceptance of the guiding conversation response based on analyzing the historical records including exception records, issue records and prediction results; and
update, by the processor, dialog nodes and conversation transition logics based on historical records.

10. The computer program product of claim 9 , wherein the program instructions executable by the processor further cause the processor to:
update, by the processor, the dialog graph based on the updated dialog nodes and transition logics.

11. The computer program product of claim 10 , wherein:
the conversation design comprises all dialog node attributes and transition logics;
a node in the dialog graph comprises the dialog node and all its dialog node attributes; and
an edge in the dialog graph comprises transition logic for any given dialog node pairs.

12. The computer program product of claim 9 , wherein the guiding conversation response is based on insertion of a temporary dialog node in the dialog graph.

13. The computer program product of claim 12 , wherein determining weights for the edges comprises:
analyzing the conversation logs;
calculating transition probabilities among dialog nodes of the dialog graph; and
weighting the edges based on the transition probabilities.

14. The computer program product of claim 9 , wherein predicting the next action comprises determining whether a match exists for a particular dialog node in the dialog graph.

15. The computer program product of claim 9 , wherein predicting the next action further comprises:
analyzing individual context information comprising: time, historical web page or mobile page visiting information, and location.

16. The computer program product of claim 15 , wherein the individual context information is used for determining potential intents and for predicting the corresponding action.

17. An apparatus comprising:
a memory configured to store instructions; and
a processor configured to execute the instructions to:
obtain information from conversation logs and a conversation design;
create a dialog graph based on the conversation design;
determine weights for edges in the dialog graph based on the information from the conversation logs and adding user input and external context information to an edge attributes set;
analyze an unrecognized user input and predict a next action based on dialog nodes in the dialog graph and historical paths;
generate a guiding conversation response based on the predicted next action;
improve the conversation design upon non-acceptance of the guiding conversation response based on analyzing the historical records including exception records, issue records and prediction results; and
update dialog nodes and conversation transition logics based on historical records.

18. The apparatus of claim 16 , wherein the processor is further configured to execute the instructions to:
the dialog graph based on the updated dialog nodes and transition logics; wherein:
the conversation design comprises all dialog node attributes and transition logics;
a node in the dialog graph comprises the dialog node and all its dialog node attributes; and
an edge in the dialog graph comprises transition logic for any given dialog node pairs.

Assignees

• IBM

Inventors

• Huang Lei
• Moore Robert J
• Ren Guangjie
• Jiang Shun

Classifications

• G06F16/3334

  Selection or weighting of terms from queries, including natural language queries · CPC title

• G06F16/3329

  Natural language query formulation · CPC title

• G06F16/9024 Primary

  Graphs; Linked lists (G06F16/9027 takes precedence) · CPC title

• G10L2015/225

  Feedback of the input speech · CPC title

• G10L15/22 Primary

  Procedures used during a speech recognition process, e.g. man-machine dialogue · CPC title