Conversation processing method, conversation management system and computer device

• US 10,162,814 B2
• Filed: 10/21/2015
• Issued: 12/25/2018
• Est. Priority Date: 10/29/2014
• Status: Active Grant

• Current Assignee: Baidu Online Network Technology (Beijing) Co., Ltd (Baidu Incorporated)
• Original Assignee: Baidu Online Network Technology (Beijing) Co., Ltd (Baidu Incorporated)
• Inventors: Jiao, Zengtao, Wang, Guanchun
• Primary Examiner(s): Albertalli, Brian L
• Application Number:
  US15/119,378
• Publication Number:
  US 20170242843A1
• Time in Patent Office: 1,161 Days
• CPC Class Codes: G06F 3/167 Audio in a user interface, e.g. using voice commands for navigating, audio feedback
• CPC Class Codes: G06F 40/174 Form filling; Merging
• CPC Class Codes: G06F 40/205 Parsing
• CPC Class Codes: G06F 40/35 Discourse or dialogue representation
• CPC Class Codes: G06F 9/46 Multiprogramming arrangements
• CPC Class Codes: G06F 9/4881 Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues

Abstract

The present invention provides a conversation processing method, a conversation management system and a computer device. The method comprises: acquiring task-related user requirement data; updating a user conversation state based on the user requirement data and a pre-configured task-related task parameter; generating at least one piece of candidate action data according to the user conversation state; and generating response text data according to the at least one piece of candidate action data. The expandability of the conversation management system is improved by customizing the conversion service related to the task in the conversation system.

Claims

1. A conversation processing method, comprising:

acquiring task-related user requirement data;

updating a user conversation state based on the user requirement data and a pre-configured task-related task parameter, the pre-configured task-related task parameter comprising multiple slots to indicate states of task parameters;

generating at least one piece of candidate action data according to the user conversation state; and

generating response text data according to the at least one piece of candidate action data,

wherein the user requirement data comprises a parse result obtained by natural language parsing of text data of a user requirement,

wherein the updating the user conversation state based on the user requirement data and the pre-configured task-related task parameter comprises;

assigning a default or a null value to the multiple slots of the pre-configured task-related task parameter based on the parse result, wherein the default value is determined from a task customization condition or a user model information;

generating a conversation parameter based on the pre-configured task-related task parameter and situation prior information; and

updating the conversation parameter according to the parse result, and

wherein generating the at least one piece of candidate action data according to the user conversation state comprises;

checking whether the user conversation state includes an ambiguous slot or an ambiguous value; and

generating a clarification candidate action in response to determining that an ambiguous slot or an ambiguous value is included in the user conversation state.

2. The method of claim 1, wherein the conversation parameter generated based on the situation prior information is expressed as a feature vector.

3. The method of claim 1, wherein the generating the at least one piece of candidate action data according to the user conversation state comprises:

checking a state of the conversation parameter according to a preset check condition, and

generating the at least one piece of candidate action data based on a result of the checking.

4. The method of claim 3, wherein the preset check condition comprises a slot default value check condition, an inter-parameter attribute dependency check condition, an inter-parameter attribute trigger check condition, a parameter value confirmation check condition and a parameter clarification condition.

5. The method of claim 1, wherein the method further comprises:

selecting preferred action data from the at least one piece of candidate action data based on the user conversation state and a decision rule, and

the generating the response text data based on the at least one piece of candidate action data comprises;

generating the response text data based on the preferred action data.

6. The method of claim 5, wherein the decision rule is one of a predetermined action priority and a preset action decision model, and

the selecting the preferred action data from the at least one piece of candidate action data comprises:

acquiring a feature function value corresponding to each of the at least one piece of candidate action data based on the action decision model, and

selecting candidate action data corresponding to a maximum of the feature function values as the preferred action data.

7. The method of claim 5, wherein the generating the response text data based on the preferred action data comprises:

acquiring parameter data required for response from a data source based on the updated conversation parameter and the preferred action data; and

generating the response text data based on the acquired parameter data and the preferred action data.

8. The method of claim 7, wherein the generating the response text data based on the preferred action data further comprises:

if the parameter data required by the preferred action data adopts a default value, filling in the parameter data with the corresponding default value.

9. The method of claim 1, wherein the updating the conversation parameter according to the parse result further comprises:

performing a data conflict detection on a value of the conversation parameter to be updated, and

emptying the value of the updated conversation parameter or sending a message for confirming whether the value of the conversation parameter to be updated is correct in response to a data conflict being detected.

10. The method of claim 9, wherein the data conflict comprises at least one of:

the value of the conversation parameter exceeding a range, subordinate relation conflict of the conversation parameter, and parameter value comparative relation conflict of the conversation parameter.

11. The method of claim 1, wherein the updating the conversation parameter based on the parse result further comprises:

calculating a value of the conversation parameter to be updated; and

updating the conversation parameter with the calculated value.

12. The method of claim 7, wherein the generating the response text data based on the acquired parameter data and the preferred action data further comprises:

performing a data conflict detection on the acquired parameter data; and

if a data conflict is detected, discarding the parameter data that has the detected data conflict.

13. The method of claim 7, wherein the generating the response text data based on the acquired parameter data and the preferred action data further comprises:

calculating values of the acquired parameter data; and

generating the response text data by using the calculated values and the preferred action data.

14. A conversation management system, comprising:

a requirement acquiring device for acquiring task-related user requirement data;

a conversation state updating device for updating a user conversation state based on the user requirement data acquired by the requirement acquiring device and a pre-configured task-related task parameter, the pre-configured task-related task parameter comprising multiple slots to indicate states of task parameters;

an action generating device for generating at least one piece of candidate action data according to the user conversation state updated by the conversation state updating device; and

a text generating device for generating response text data according to the at least one piece of candidate action data generated by the action generating device,

wherein the user requirement data comprises a parse result obtained by natural language parsing of text data of a user requirement,

wherein the conversation state updating device is configured to;

assign a default or a null value to the multiple slots of the pre-configured task-related task parameter based on the parse result, wherein the default value is determined from a task customization condition or a user model information;

generate a conversation parameter based on the pre-configured task-related task parameter and situation prior information; and

update the conversation parameter according to the parse result, and

wherein the action generating device is configured to;

check whether the user conversation state includes an ambiguous slot or an ambiguous value, and

generate a clarification candidate action in response to determining that an ambiguous slot or an ambiguous value is included in the user conversation state.

15. The system of claim 14, wherein the conversation state updating device is further configured to express the conversation parameter generated based on the situation prior information as a feature vector.

16. The system of claim 14, wherein the action generating device is configured for checking a state of the conversation parameter based on a preset check condition, and generating the at least one piece of candidate action data based on a result of the checking.

17. The system of claim 16, wherein the preset check condition comprises a slot default value check condition, an inter-parameter attribute dependency check condition, an inter-parameter attribute trigger check condition, a parameter value confirmation check condition and a parameter clarification condition.

18. The system of claim 14, wherein the system further comprises an action decision device, which is configured for

selecting preferred action data from the at least one piece of candidate action data based on the user conversation state and a decision rule, and

the text generating device is configured for generating the response text data based on the preferred action data.

19. The system of claim 18, wherein

the decision rule is one of a predetermined action priority and a preset action decision model, and

the action decision device is configured for acquiring a feature function value corresponding to each of the at least one piece of candidate action data based on the action decision model, and selecting candidate action data corresponding to a maximum of the feature function values as the preferred action data.

20. A computer device, comprising:

one or more processors;

a memory; and

one or more programs, the one or more programs being stored in the memory and comprising instructions configured to be executed by the one or more processors to implement a conversation processing method, the method comprising;

acquiring task-related user requirement data;

updating a user conversation state based on the user requirement data and a pre-configured task-related task parameter, the pre-configured task-related task parameter comprising multiple slots to indicate states of task parameters;

generating at least one piece of candidate action data according to the user conversation state; and

generating response text data according to the at least one piece of candidate action data,

wherein the user requirement data comprises a parse result obtained by natural language parsing of text data of a user requirement,

wherein the updating the user conversation state based on the user requirement data and the pre-configured task-related task parameter comprises;

assigning a default or a null value to the multiple slots of the pre-configured task-related task parameter based on the parse result, wherein the default value is determined from a task customization condition or a user model information;

generating a conversation parameter based on the pre-configured task-related task parameter and situation prior information; and

updating the conversation parameter according to the parse result, and

wherein generating the at least one piece of candidate action data according to the user conversation state comprises;

checking whether the user conversation state includes an ambiguous slot or an ambiguous value; and

generating a clarification candidate action in response to a determination that an ambiguous slot or an ambiguous value is included in the user conversation state.