Dialogue system for user requests controlling network devices by FSM with plan language

• US 7,702,607 B2
• Filed: 04/05/2005
• Issued: 04/20/2010
• Est. Priority Date: 04/06/2004
• Status: Expired due to Fees

• Current Assignee: Sony Deutschland GmbH (Sony Group Corp.)
• Original Assignee: Sony Deutschland GmbH (Sony Group Corp.)
• Inventors: Torge, Sunna, Emele, Martin
• Primary Examiner(s): Vincent; David R
• Assistant Examiner(s): CHANG, LI WU
• Application Number:
  US11/099,084
• Publication Number:
  US 20050234726A1
• Time in Patent Office: 1,841 Days
• Field of Search: 706/45, 706/46
• US Class Current: 706/46
• CPC Class Codes: G06N 5/04 Inference or reasoning models

Abstract

A Method of operating a dialogue system for serving complex user requests by appropriately controlling a network of devices (D1, . . . , Dn) and/or applications, wherein each device (D1, . . . , Dn) and/or application and its functionality are modeled by a finite state machine (FSM-1, FSM-2, FSM-3), which is using a plan language, wherein a reasoning component (1) is used in order to determine a plan (P) using said plan language for serving a received complex user request, wherein said reasoning component (1) is adapted for determining said plan (P) using said finite state machines (FSM-1, FSM-2, FSM-3), of said devices (D1, . . . , Dn) and/or applications, wherein a plan processing module (2) is used in order to perform a determined plan (P) for serving a received user request, and wherein said plan language is designed in order to enable the determination of conditional plans by said reasoning component (1), in particular by providing conditional planning commands.

Claims

1. Method for operating a dialogue system for serving complex user requests by appropriately controlling a network of devices and/or applications,

wherein each device and/or application is modeled by a finite state machine, said finite state machine being descriptive for the functions and states of a respective device and/or application,

wherein transitions of said finite state machine are annotated with statements of a plan language,

the method comprising:

receiving a user request by a dialogue manager adapted to receive a user request from a user;

sending the user request to a reasoning component of a function planning module; and

determining a determined plan by the reasoning component based on the received user request, said reasoning component being configured to use said finite state machines in order to determine said determined plan, said determined plan being based on said plan language,

wherein the function planning module includes a plan processing module which is used in order to perform said determined plan for serving the received user request,

wherein said plan language is designed in order to enable the determination of conditional plans by said reasoning component, by providing conditional planning commands,

wherein said conditional planning commands include at least one EXPECT-statement,

wherein said reasoning component is configured to determine a conditional plan as determined plan by using said at least one EXPECT-statement which corresponds to an expected feedback of a device and/or application,

wherein said plan processing module is configured to receive a current feedback from a device and/or an application when said at least one EXPECT-statement within said conditional plan is executed, wherein if said current feedback corresponds to said expected feedback, the plan processing module is configured to further execute said conditional plan, and else if said current feedback does not correspond to said expected feedback, the plan processing module is configured to execute an alternative plan,

wherein said plan language comprises a WAIT-statement and said reasoning component plans at least one further plan in case said plan processing module is executing or previously has executed a WAIT-statement and is thus in a waiting state,

wherein said at least one further plan is executed by said plan processing module, such that said at least one further plan is executed in parallel to said determined plan.

2. Method according to claim 1, wherein said plan processing module is adapted for realizing one subprocess or any combination of sub-processes of the following group of sub-processes consisting of:

serving different tasks in parallel,

scheduling tasks at given certain times, and

interrupting tasks with respect to user and/or application interrupts.

19. Computer readable storage medium encoded with instructions, which when executed by computer, causes the computer to implement the method of claim 1.

3. Method for operating a dialog system for serving complex user requests by appropriately controlling a network of devices and/or applications:

wherein each device and/or application is modeled by a finite state machine, said finite state machine being descriptive for the functions and states of a respective device and/or application,

wherein transitions of said finite state machine are annotated with statements of a plan language,

the method comprising;

receiving a user request by a dialogue manager configured to receive a user request from a user;

sending the user request to a reasoning component of a function planning module; and

determining a determined plan by the reasoning component based on the received user request, said reasoning component being adapted for using said finite state machines in order to determine said determined plan, said determined plan being based on said plan language,

wherein the function planning module includes a plan processing module which is used in order to perform said determined plan for serving the received user request, and

wherein said plan processing module is configured to realize one sub-process or any combination of sub-processes of the following group of sub-processes;

serving different tasks in parallel,

scheduling tasks at given certain times, and

interrupting tasks with respect to user and/or application interrupts,

wherein said plan language is designed in order to enable the determination of conditional plans by said reasoning component, by providing conditional planning commands,

wherein said conditional planning commands include at least one EXPECT-statement,

wherein said reasoning component is configured to determine a conditional plan as determined plan by using said at least one EXPECT-statement which corresponds to an expected feedback of a device and/or application,

wherein said plan processing module is configured to receive a current feedback from a device and/or an application when said at least one EXPECT-statement within said conditional plan is executed, wherein if said current feedback corresponds to said expected feedback, the plan processing module is configured to further execute said conditional plan, and else if said current feedback does not correspond to said expected feedback, the plan processing module is configured to execute an alternative plan,

wherein said plan language comprises a WAIT-statement and said reasoning component plans at least one further plan in case said plan processing module is executing or previously has executed a WAIT-statement and is thus in a waiting state,

wherein said at least one further plan is executed by said plan processing module, such that said at least one further plan is executed in parallel to said determined plan.

4. Method according to claim 3, wherein said method further comprises:

a planning stage, wherein said determined plan or at least one further plan is determined or said determined plan is modified by said reasoning component,

an executing stage, wherein said determined plan and/or said at least one further plan is executed by said plan processing module, and

at least one switching step for switching back and forth between said planning stage and said executing stage, in particular when processing a conditional plan or serving different tasks in parallel.

5. Method according to claim 4, wherein said plan language comprises a REPEAT_WHILE_PENDING-statement that describes an undefined number of WAIT-statements, thus realizing a determined plan of an undefined length comprising an unknown number of statements.

6. Method according to claim 3, wherein said alternative plan is loaded by said plan processing module, or said alternative plan is newly generated by said reasoning component.

7. Method according to claim 5, wherein said alternative plan is newly generated by said reasoning component based on a current state of said dialogue system and/or based on current states of said devices and/or applications.

8. Method according to claim 3, wherein said reasoning component and/or said plan processing module comprise an internal state data base with a data base entry for each device, said data base entry being descriptive for a current state of a respective device, said plan language comprises a GOAL-statement describing a goal state of a certain device and/or application and a REACHED-statement for setting said data base entry of said certain device and/or application to said goal state, wherein, when said REACHED-statement is processed by said plan processing module, the data base entry for said certain device and/or application within the internal state data base is updated and set to said goal state.

9. Method according to claim 8, wherein said determined plan comprises said GOAL-statement followed by following statements, followed by said REACHED-statement, said GOAL-statement is processed by said plan processing module, and in case said certain device and/or application is already in said goal state, said following statements are skipped and said REACHED-statement is processed directly after said GOAL-statement, and in case said certain device and/or application is not in said goal state, said following statements are processed, and said REACHED-statement is processed by said plan processing module thereby setting said data base entry for said certain device and/or application to said goal state.

10. Method according to claim 3, wherein certain functionalities are described by virtual devices that are modeled by a respective finite state machine.

11. Method according to claim 10, wherein said virtual devices comprise a scheduling device for providing a scheduling function within a determined plan.

12. Method according to claim 11, wherein said scheduling function is provided to said devices and/or applications of said network and thus enables the scheduling of tasks provided by said devices and/or applications of said network and/or the scheduling of a determined plan.

13. Method according to claim 10, wherein said virtual devices comprise a stopping device for providing a stopping function within a determined plan.

14. Method according to claim 13, wherein said stopping function is provided to said devices and/or applications of said network and thus enables the stopping of tasks provided by said devices and/or applications of said network and/or the stopping of a determined plan.

15. Method according to claim 3, wherein said plan language and said determined plan comprises exclusively any or any combination exclusively of the following statements:

ACTION;

instructing a device to perform a task;

WAIT;

modeling that a device and/or application is in a waiting state;

REPEAT_WHILE_PENDING;

modeling an unknown number of WAIT-statements;

EXPECT;

in case an expected feedback is received from a device and/or application, said determined plan is continued to be executed, otherwise an alternative plan is executed by said plan processing module;

GOAL;

in case a device is in a goal state that is indicated by said GOAL-statement, following statements are skipped until a REACHED-statement is reached, and in case said device and/or application is not in said goal state, the following statements are processed;

REACHED;

updating an internal state data base with the current state of a respective device and/or application;

thus enabling a simple description language regarding expressive power with a low complexity.

16. Method according to claim 3 wherein in case said reasoning component determines several possible plans for serving a received complex user request, said reasoning component selects said determined plan from said several possible plans depending on cost aspects for the user.

17. Method according to claim 3, wherein in case said reasoning component determines several possible plans for serving a received complex user request, said reasoning component selects said determined plan based on a selection of the user and/or based on predetermined preferences.

18. Dialogue system for serving complex user requests by appropriately controlling a network of devices and/or applications, comprising:

a device and/or application that is modeled by a finite state machine, said finite state machine being descriptive for the functions and states of a respective device and/or application,

wherein transitions of said finite state machine are annotated with statements of a plan language;

means for receiving a user request by a dialogue manager adapted to receive a user request from a user;

means for sending the user request to a reasoning component of a function planning module; and

means for determining a determined plan by the reasoning component based on the received user request, said reasoning component being configured to use said finite state machines in order to determine said determined plan, said determined plan being based on said plan language,

wherein the function planning module includes a plan processing module which is used in order to perform said determined plan for serving the received user request,

wherein said plan language is designed in order to enable the determination of conditional plans by said reasoning component, by providing conditional planning commands,

wherein said conditional planning commands include at least one EXPECT-statement,

wherein said reasoning component is configured to determine a conditional plan as determined plan by using said at least one EXPECT-statement which corresponds to an expected feedback of a device and/or application, and

wherein said plan processing module is configured to receive a current feedback from a device and/or an application when said at least one EXPECT-statement within said conditional plan is executed, wherein if said current feedback corresponds to said expected feedback, the plan processing module is configured to further execute said conditional plan, and else if said current feedback does not correspond to said expected feedback, the plan processing module is configured to execute an alternative plan,

wherein said plan language comprises a WAIT-statement and said reasoning component plans at least one further plan in case said plan processing module is executing or previously has executed a WAIT-statement and is thus in a waiting state,

wherein said at least one further plan is executed by said plan processing module, such that said at least one further plan is executed in parallel to said determined plan.