Actuation system for an agent oriented architecture

• US 7,353,176 B1
• Filed: 12/20/2002
• Issued: 04/01/2008
• Est. Priority Date: 12/20/2001
• Status: Active Grant

• Current Assignee: iAnywhere Solutions Incorporated (SAP SE)
• Original Assignee: iAnywhere Solutions Incorporated (SAP SE)
• Inventors: Joshi, Aparna, Yuen, Daniel Charles, Baray, Cristobal H., Hodjat, Babak
• Primary Examiner(s): ABEBE, DANIEL DEMELASH
• Application Number:
  US10/327,440
• Time in Patent Office: 1,929 Days
• Field of Search: 704/275, 704/257
• US Class Current: 704/275
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

Roughly described, the invention addresses the above problems by separating an actuation subsystem from the natural language interpretation system. The NLI develops “interpretation result commands” in response to user input, and transmits them to the actuation subsystem using a predefined interpretation result command format that is independent of the requirements of the particular back-end application. The actuation subsystem, which is the only component that is specific to the back-end application, converts the interpretation result command into one or more “application commands” and communicates it (them) to the back-end application in the form required by the specific back-end application. In some embodiments the actuation subsystem also can take results from a back-end application in application-dependent form, and convert it to a common predefined internal format that is application-independent.

Claims

1. A user interaction method, comprising the steps of:

instantiating an actuation module having a back-end interface arranged to communicate with a first back-end application; and

instantiating a natural language interpreter having an input for receiving token sequences generated from user input, and an output for outputting interpretation result commands in an application-independent form toward said actuation module in response to said token sequences;

said actuation module receiving a first one of said interpretation result commands, translating said first interpretation result command into a set of at least one application-specific application command for said first back-end application, and forwarding said at least one application command toward said first back-end application, and

said actuation module receiving a command result from said first back-end application, translating said command result into a common interaction format, and forwarding said command result toward said user.

2. A method according to claim 1, further comprising the steps of:

said natural language interpreter issuing a clarification request in response to one of said token sequences and forwarding said clarification request toward said user.

3. A method according to claim 1, wherein said first back-end application supports at least one member of the group consisting of email management and contacts management.

4. A method according to claim 1, further comprising the steps of:

in conjunction with said actuation module translating said first interpretation result command into a set of at least one application-specific application command for said first back-end application, said actuation module developing a set of at least one application-specific undo command for said first back-end application, the set of at least one application-specific undo command being effective to undo the set of at least one application-specific application command; and

in response to user input after the step of said actuation module forwarding said command result toward said user, said actuation module forwarding said at least one application-specific undo command toward said first back-end application.

5. A user interaction method, comprising the steps of:

instantiating an actuation module having a back-end interface arranged to communicate with a first back-end application; and

instantiating a natural language interpreter having an input for receiving token sequences generated from user input, and an output for outputting interpretation result commands in an application-independent form toward said actuation module in response to said token sequences;

said actuation module receiving a first one of said interpretation result commands, translating said first interpretation result command into a set of at least one application-specific application command for said first back-end application, and forwarding said at least one application command toward said first back-end application, and

said natural language interpreter issuing a clarification request in response to one of said token sequences and forwarding said clarification request toward said user.

6. A user interaction method, comprising the steps of:

instantiating an actuation module having a first back-end interface arranged to communicate with a first back-end application and a second back-end interface arranged to communicate with a second back-end application; and

instantiating a natural language interpreter having an input for receiving token sequences generated from user input, and an output for outputting interpretation result commands in an application-independent form toward said actuation module in response to said token sequences;

said actuation module receiving a first one of said interpretation result commands, translating said first interpretation result command into a set of at least one application-specific application command for said first back-end application, and forwarding said at least one application command toward said first back-end application,

and said actuation module receiving a second one of said interpretation result commands, translating said second interpretation result command into a set of at least one second application command for said second back-end application, and forwarding said at least one second application command toward said second back-end application.

7. A method according to claim 6, further comprising the steps of:

said actuation module receiving a second command result from said second back-end application, translating said second command result into said common interaction format, and forwarding said second command result toward said user.

8. A user interaction method, comprising the steps of:

instantiating an actuation module having a back-end interface arranged to communicate with a first back-end application; and

instantiating a natural language interpreter having an input for receiving token sequences generated from user input, and an output for outputting interpretation result commands in an application-independent form toward said actuation module in response to said token sequences,

said actuation module receiving a first one of said interpretation result commands, translating said first interpretation result command into a set of at least one application-specific application command for said first back-end application, and forwarding said at least one application command toward said first back-end application,

wherein said natural language interpreter comprises;

a plurality of agents arranged in a network, each of said agents having at least one interpretation policy having a policy condition and a policy action associated with the condition, each of said interpretation policies making a claim on at least part of a token sequence provided to the agent if the condition of the interpretation policy is satisfied, and executing the associated policy action in response to determination that the claim is at least part of a winning claim.

9. A method according to claim 8, wherein a particular one of said agents has a particular interpretation policy having a policy condition that refers to a downchain one of said agents, any claim made by said particular policy being dependent upon any claims made by said downchain agent.

10. A method according to claim 8, wherein a particular one of said agents has a particular interpretation policy having a particular policy condition that refers to a downchain one of said agents, said particular interpretation policy further having a particular policy action associated with said particular policy condition,

wherein said particular policy action includes at least partial delegation to said downchain agent.

11. A user interaction method, comprising the steps of:

instantiating an actuation module having a back-end interface arranged to communicate with a first back-end application;

instantiating a natural language interpreter having an input for receiving token sequences generated from user input, and an output for outputting interpretation result commands toward said actuation module in response to said token sequences;

said natural language interpreter outputting a first interpretation result command toward said first back-end application in response to a first one of said token sequences;

said first back-end application providing a first response to said first interpretation result command in an application-specific form; and

said actuation module translating said first response into an application-independent form, and communicating said first response toward said user.

12. A method according to claim 11, further comprising the step of instantiating an I/O subsystem,

further comprising the step of said I/O subsystem generating said first token sequence in response to first user input,

and wherein said step of said actuation module communicating said first response toward said user comprises the steps of:

said actuation module communicating said first response to said I/O subsystem, bypassing said natural language interpreter; and

said I/O subsystem forwarding said response toward said user.

13. A method according to claim 11, further comprising the steps of:

said natural language interpreter outputting a second interpretation result command toward a second back-end application in response to a second one of said token sequences;

said second back-end application providing a second response to said second interpretation result command in an application-specific form; and

said actuation module translating said second response into said application-independent form, and communicating said second response toward said user.

14. A method according to claim 11, for use with a plurality of back-end applications including said first application, wherein said step of said natural language interpreter outputting a first interpretation result command toward said back-end application comprises the steps of:

said natural language interpreter outputting said first interpretation result command to said actuation module;

said actuation module developing a plurality of application commands in response to said first interpretation result command; and

said actuation module communicating each of said application commands to a corresponding destination one of said plurality of back-end applications, said destination applications including at least said first back-end application and a second back-end application,

said second back-end application providing a second response to said first interpretation result command in an application-specific form; and

said actuation module combining said second response with said first response for communicating toward said user in said application-independent form.

15. A method according to claim 14, wherein said step of said actuation module communicating each of said application commands to a corresponding destination one of said plurality of back-end applications comprises the step of said actuation module communicating each of said application commands to the corresponding destination one of said plurality of back-end applications in a form specific to the destination application.

16. A method according to claim 11, wherein said natural language interpreter comprises a distributed parser of natural language input.

17. A method according to claim 11, wherein said natural language interpreter comprises an agent network.

18. A method according to claim 11, wherein said natural language interpreter comprises:

a plurality of agents arranged in a network, each of said agents having at least one interpretation policy having a policy condition and a policy action associated with the condition, each of said interpretation policies making a claim on at least part of a token sequence provided to the agent if the condition of the interpretation policy is satisfied, and executing the associated policy action in response to determination that the claim is at least part of a winning claim.

19. A method according to claim 18, wherein a particular one of said agents has a particular interpretation policy having a policy condition that refers to a downchain one of said agents, any claim made by said particular policy being dependent upon any claims made by said downchain agent.

20. A method according to claim 18, wherein a particular one of said agents has a particular interpretation policy having a particular policy condition that refers to a downchain one of said agents, said particular interpretation policy further having a particular policy action associated with said particular policy condition,

wherein said particular policy action includes at least partial delegation to said downchain agent.

21. A method according to claim 11, wherein said agent network prepares said commands in dependence upon satisfaction, in said token sequences, of one or more policy conditions which make reference to one or more keywords in said predetermined set of keywords.

22. User interactive apparatus comprising:

a natural language interpreter having an input for receiving token sequences from a user interaction subsystem and an output for outputting interpretation result commands in response to said token sequences; and

an actuation subsystem in communication with first and second back-end applications, said actuation subsystem communicating first and second application commands to respectively said first and second back-end applications, both in response to a single one of said interpretation result commands.

23. Apparatus according to claim 22, wherein said natural language interpreter comprises a distributed parser of natural language input.

24. Apparatus according to claim 22, wherein said natural language interpreter comprises an agent network.

25. Apparatus according to claim 22, wherein said natural language interpreter comprises:

a plurality of agents arranged in a network, each of said agents having at least one interpretation policy having a policy condition and a policy action associated with the condition, each of said interpretation policies making a claim on at least part of a token sequence provided to the agent if the condition of the interpretation policy is satisfied, and executing the associated policy action in response to determination that the claim is at least part of a winning claim.

26. Apparatus according to claim 25, wherein a particular one of said agents has a particular interpretation policy having a policy condition that refers to a downchain one of said agents, any claim made by said particular policy being dependent upon any claims made by said downchain agent.

27. Apparatus according to claim 25, wherein a particular one of said agents has a particular interpretation policy having a particular policy condition that refers to a downchain one of said agents, said particular interpretation policy further having a particular policy action associated with said particular policy condition,

wherein said particular policy action includes at least partial delegation to said downchain agent.

28. Apparatus according to claim 22, wherein said agent network prepares said interpretation result commands in dependence upon satisfaction, in said token sequences, of one or more policy conditions which make reference to one or more keywords in said predetermined set of keywords.

29. Apparatus according to claim 22, wherein said actuation subsystem further communicates command results from each of said first and second back-end applications to said user interaction subsystem.

30. Apparatus according to claim 29, wherein said actuation subsystem converts said command results to a common interaction format before communicating them to said user interaction subsystem.

31. Apparatus according to claim 30, wherein said natural language interpreter further has an output for providing clarification requests toward said user.

32. Apparatus according to claim 31, wherein said natural language interpreter communicates said clarification requests toward said user via said user interaction subsystem in said common interaction format.

33. Apparatus according to claim 22, wherein said first back-end application supports at least one member of the group consisting of email management and contacts management.

34. User interactive apparatus comprising:

a natural language interpreter having an input for receiving token sequences from a user interaction subsystem and an output for outputting interpretation result commands in response to said token sequences; and

an actuation subsystem in communication with first and second back-end applications said actuation subsystem communicating first and second application commands to respectively said first and second back-end applications in response to said interpretation result commands,

wherein said actuation subsystem communicates said first application command to said first back-end application in response to a first one of said interpretation result commands and communicates said second application command to said second back-end application in response to a second one of said interpretation result commands.

35. User interactive apparatus comprising:

a natural language interpreter having an input for receiving user input from a user and an output for outputting interpretation result commands toward first and second back-end applications in response to said user input; and

an actuation subsystem having interfaces to each of said first and second back-end applications, said actuation subsystem being arranged to receive command results from both said first and second applications, and to convert said command results to a common interaction format, and to communicate them in said common interaction format toward said user.

36. Apparatus according to claim 35, wherein said actuation subsystem communicates said command results in said common interaction format toward said user via a path that bypasses said natural language interpreter.

37. Apparatus according to claim 35, wherein said natural language interpreter comprises a distributed parser of natural language input.

38. Apparatus according to claim 35, wherein said natural language interpreter comprises an agent network.

39. A user interaction method, for use with a first back-end application, comprising the steps of:

developing an interpretation result command in an application-independent form in response to token sequences generated from user input;

developing a set of at least one application-specific undo command for said first back-end application in dependence upon said application-independent interpretation result command, the set of at least one application-specific undo command being effective to undo the interpretation result command; and

in response to subsequent user input, forwarding said at least one application-specific undo command toward said first back-end application.

40. A method according to claim 39, further comprising the steps of:

translating said application-independent interpretation result command into a set of at least one application-specific application command for said first back-end application; and

forwarding said at least one application command toward said first back-end application.