Systems and methods for processing natural language queries

• US 20050289124A1
• Filed: 06/29/2004
• Published: 12/29/2005
• Est. Priority Date: 06/29/2004
• Status: Active Grant

• Current Assignee: SAP SE
• Original Assignee: SAP AG (SAP SE)
• Inventors: Klein, Jacob A., Kaiser, Matthias, Jiang, Shan, Vogler, Hartmut
• Granted Patent: US 7,720,674 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/24522 Translation of natural language queries to structured queries

Abstract

Methods and systems are provided for processing natural language queries. Such methods and systems may receive a natural language query from a user and generate corresponding semantic tokens. Information may be retrieved from a knowledge base using the semantic tokens. Methods and systems may leverage an interpretation module to process and analyze the retrieved information in order to determine an intention associated with the natural language query. Methods and systems may leverage an actuation module to provide results to the user, which may be based on the determined intention.

Claims

1. A method for processing natural language queries comprising:

obtaining a natural language query from a user;

generating at least one semantic token from the natural language query;

identifying data in a knowledge base using the at least one semantic token;

interpreting the identified data based on an intention associated with the user; and

actuating the interpreted data.

2. The method of claim 1, wherein obtaining the natural language query comprises obtaining at least one of an alphanumeric character, an audio signal, and a visual signal.

3. The method of claim 1, wherein identifying data in a knowledge base comprises searching a structured data archive using the semantic token.

4. The method of claim 1, wherein actuating the interpreted data comprises translating the interpreted data into at least one system-actionable command.

5. The method of claim 1, wherein actuating the interpreted data comprises providing the interpreted data to the user.

6. The method of claim 1, the method further comprising:

identifying at least one term having a semantic relationship with the at least one semantic token.

7. The method of claim 6, wherein identifying data in a knowledge base comprises identifying data in a knowledge base using the at least one semantic token and the at least one semantically-related term.

8. The method of claim 7, wherein the at least one semantically-related term comprises a synonym.

9. The method of claim 1, wherein interpreting the identified data comprises:

recognizing an uncertainty in the identified data; and

resolving the uncertainty based on the user intention.

10. The method of claim 9, wherein recognizing an uncertainty comprises recognizing that the identified data relates to a plurality of contexts that potentially relate to the user intention.

11. The method of claim 9, wherein recognizing an uncertainty comprises recognizing that the identified data relates to a plurality of data types that potentially relate to the user intention.

12. The method of claim 9, wherein recognizing an uncertainty comprises recognizing a deficiency in the natural language query.

13. The method of claim 9, wherein resolving the uncertainty comprises establishing a dialog with the user to resolve the uncertainty.

14. The method of claim 9, wherein resolving the uncertainty comprises applying a preset rule.

15. The method of claim 9, wherein resolving the uncertainty comprises:

establishing a policy for the user;

applying the policy to resolve the uncertainty;

receiving feedback from the user relating to the application of the policy; and

re-configuring the policy based on the feedback.

16. A method for processing natural language queries comprising:

obtaining a natural language query from a user;

generating at least one semantic token from the natural language query;

identifying data in a knowledge base using the at least one semantic token;

determining an intention associated with the user based on the identified data; and

providing information that is relevant to the natural language query to the user based on the determined intention.

17. The method of claim 16, wherein determining an intention comprises recognizing an uncertainty in the identified data.

18. The method of claim 17, wherein recognizing an uncertainty comprises recognizing that the identified data relates to a plurality contexts that potentially relate to the natural language query.

19. The method of claim 17, wherein recognizing an uncertainty comprises recognizing that the identified data relates to a plurality of data types that potentially relate to the natural language query.

20. The method of claim 17, wherein recognizing an uncertainty comprises recognizing a deficiency in the natural language query.

21. The method of claim 17, wherein determining an intention comprises resolving the uncertainty.

22. The method of claim 21, wherein resolving the uncertainty comprises, establishing a dialog with the user.

23. The method of claim 21, wherein resolving the uncertainty comprises applying a preset rule.

24. The method of claim 21, wherein resolving the uncertainty comprises:

establishing a policy for the user;

applying the policy to resolve the uncertainty;

receiving feedback from the user relating to the application of the policy; and

re-configuring the policy based on the feedback.

25. The method of claim 16, the method further comprising:

identifying at least one term having a semantic relationship with the at least one semantic token generated from the natural language query.

26. The method of claim 25, wherein identifying data in a knowledge base comprises identifying data in a knowledge base using the at least one semantic token and the at least one semantically-related term.

27. The method of claim 26, wherein the at least one semantically-related term comprises a synonym.

28. A method for processing natural language queries comprising:

obtaining a natural language query from a source;

retrieving data that is potentially relevant to the natural language query from a knowledge base;

determining an intention associated with the natural language query; and

processing the potentially relevant data in accordance with the intention so as to identify actually relevant data from the potentially relevant data; and

providing the actually relevant data to the source.

29. The method of claim 28, wherein obtaining the natural language query comprises obtaining at least one of an alphanumeric character, an audio signal, and a visual signal from a user.

30. The method of claim 28, wherein retrieving the potentially relevant data comprises searching a structured data archive using at least one semantic token derived from the natural language query.

31. The method of claim 28, wherein determining the intention comprises interacting with a user to determine the intention.

32. The method of claim 28, wherein processing the potentially relevant data comprises applying a preset rule in accordance with the intention.

33. The method of claim 28, wherein processing the potentially relevant data comprises:

establishing a policy for the user in accordance with the intention;

applying the policy to identify the actually relevant data from the potentially relevant data;

receiving feedback from the user relating to application of the policy; and

re-configuring the policy based on the feedback.

34. A method for processing natural language queries, comprising:

obtaining a natural language query from a user;

generating at least one semantic token from the natural language query;

identifying data in a knowledge base using the at least one semantic token;

identifying an uncertainty associated with the identified data and the natural language query;

determining an intention associated with the user based on the identified data; and

resolving the identified uncertainty based on the determined intention.

35. The method of claim 34, wherein identifying an uncertainty comprises identifying that additional information is required from the user.

36. The method of claim 34, wherein identifying an uncertainty comprises identifying that the data identified in the knowledge base relates to a plurality of contexts.

37. The method of claim 34, wherein identifying an uncertainty comprises identifying that the data identified in the knowledge base relates to a plurality of data types.

38. The method of claim 34, wherein identifying an uncertainty comprises identifying that the natural language query includes a series of dependent commands and identifying that information associated with a sequence of the commands is required.

39. A system for processing natural language queries, comprising:

means for obtaining a natural language query from a user;

means for generating at least one semantic token from the natural language query;

means for identifying data in a knowledge base using the at least one semantic token;

means for determining an intention associated with the user based on the identified data; and

means for providing information that is relevant to the natural language query to the user based on the determined intention.

40. A system for processing natural language queries, comprising:

means for obtaining a natural language query from a user;

means for generating at least one semantic token from the natural language query;

means for identifying data in a knowledge base using the at least one semantic token;

means for identifying an uncertainty associated with the identified data and the natural language query;

means for determining an intention associated with the user based on the identified data; and

means for resolving the identified uncertainty based on the determined intention.

41. A natural language query processing system, comprising:

an interface module configured to receive a natural language query;

a tokenizing module configured to generate at least one semantic token based on the received natural language query;

a searching module configured to retrieve information from a knowledge base using the at least one semantic token;

an interpretation module configured to;

identify an uncertainty associated with the retrieved information, and

process the retrieved information so as to resolve the uncertainty based on an intention associated with the received natural language query; and

an actuation module configured to translate the processed information into a system-actionable command.

42. The system of claim 41, wherein the interface modules is configured to receive the natural language query from at least one of image capture device, an audio capture device, a keyboard, a mouse, a touch screen.

43. The system of claim 41, wherein the interface module is configured to receive the natural language query from a data processing system via a network.

44. The system of claim 41, wherein the tokenizing module is further configured to discard at least one element of the natural language query.

45. The system of claim 41, wherein the searching module is configured to retrieve the information from a structured data archive.

46. The system of claim 41, wherein the searching module is further configured to identify at least one term having a semantic relationship with the at least one semantic token.

47. The system of claim 46, wherein the searching module is configured to retrieve information from the knowledge base using the at least one semantic token and the semantically-related term.

48. The system of claim 47, wherein the semantically-related term comprises one of a synonym and a hypernym.

49. The system of claim 41, wherein the interpretation module is configured to resolve the uncertainty based on an intention associated with the received natural language query by retrieving information from a user.

50. The system of claim 41, wherein the interpretation module identifies the uncertainty by recognizing that the retrieved information relates to a plurality contexts.

51. The system of claim 41, wherein the interpretation module identifies the uncertainty by recognizing that the retrieved information relates to a plurality of data types.

52. The system of claim 41, wherein the interpretation module identifies the uncertainty by recognizing a deficiency in the natural language query.

53. The system of claim 41, wherein the interpretation module resolves the uncertainty by applying a preset rule.

54. The system of claim 41, wherein the interpretation module resolves the uncertainty by:

establishing a policy;

applying the policy to resolve the uncertainty;

55. The system of claim 54, wherein the interpretation module is configured to:

receive feedback relating to the application of the policy; and

re-configure the policy based on the feedback.

56. A natural language query processing system, comprising:

an interface module configured to receive a natural language query;

a tokenizing module configured to generate at least one semantic token based on the received natural language query;

a searching module configured to retrieve information from a knowledge base using the at least one semantic token;

an interpretation module configured to;

determine an intention associated with the received natural language query, and

process the retrieved information in accordance with the intention; and

an actuation module configured to provide the processed information to a user.

57. The system of claim 56, wherein the interface modules is configured to receive the natural language query from at least one of image capture device, an audio capture device, a keyboard, a mouse, a touch screen.

58. The system of claim 56, wherein the interface module is configured to receive the natural language query from a data processing system via a network.

59. The system of claim 56, wherein the tokenizing module is further configured to discarding at least one elements of the natural language query.

60. The system of claim 56, wherein the searching module is configured to retrieve the information from a structured data archive.

61. The system of claim 56, wherein the searching module is further configured to identify at least one term having a semantic relationship with the at least one semantic token.

62. The system of claim 61, wherein the searching module is configured to retrieve information from the knowledge base using the at least one semantic token and the semantically-related term.

63. The system of claim 62, wherein the semantically-related term comprises one or synonym and a hypernym.

64. The system of claim 56, wherein the interpretation module is configured to determine the intention by communication with a user.

65. The system of claim 56, wherein the interpretation module determines the intention by:

recognizing an uncertainty associated with the retrieved information and the natural language query; and

retrieving clarifying data from a user in order to resolve the ambiguity.

66. The system of claim 65, wherein the interpretation module recognizes the uncertainty by recognizing that the retrieved information relates to a plurality contexts.

67. The system of claim 65, wherein the interpretation module recognizes the uncertainty by recognizing that the retrieved information relates to a plurality of data types.

68. The system of claim 65, wherein the interpretation module recognizes the uncertainty by recognizing a deficiency in the natural language query.

69. The system of claim 65, wherein the interpretation module retrieves clarifying data from a user by establishing a dialog with the user via a data processing system.

70. The system of claim 56, wherein the interpretation module processes the retrieved information by applying a preset rule.

71. The system of claim 56, wherein the interpretation module processes the retrieved information by:

establishing a policy;

applying the policy to resolve an uncertainty associated with the retrieved data and the natural language query;

receiving feedback relating to the application of the policy; and

re-configuring the policy based on the feedback.

72. A computer-readable medium containing instructions for controlling a computer system coupled to a network to perform a method, the computer system having a processor for executing the instructions, the method comprising:

obtaining a natural language query from a source;

retrieving data that is potentially relevant to the natural language query from a knowledge base;

determining an intention associated with the natural language query; and

processing the potentially relevant data in accordance with the intention so as to separate the potentially relevant data into actually relevant data and actually irrelevant data; and

providing the actually relevant data to the source.