System and method for electronic communication management

• US 7,099,855 B1
• Filed: 01/03/2001
• Issued: 08/29/2006
• Est. Priority Date: 01/13/2000
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Cohen, Dani, Hajaj, Nissan, Magdalen, Josemina, Nelken, Yoram, Bluger, Michael, Nelken, Israel
• Primary Examiner(s): Hirl, Joseph P.
• Application Number:
  US09/754,179
• Time in Patent Office: 2,064 Days
• Field of Search: 706/50, 706/909, 706/14, 706/21, 706/47, 706/46, 706/55, 709/206, 709/202, 704/9, 704/236, 704/255, 382/159, 707/10, 707/3, 705/26, 717/104, 717/105
• US Class Current: 706/50
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition
• CPC Class Codes: G06Q 10/00 Administration; Management
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

A system and method for electronic communication management comprises a universal data model, a modeling engine, and an adaptive knowledge base. The modeling engine includes a natural language processor and a statistical modeler. A communication is translated from its native format into the universal data model. The modeling engine determines the intent of the communication using the natural language processor and the statistical modeler. A response is generated, either automatically or by an agent. An audit module analyzes each response and provides feedback to the modeling engine and the adaptive knowledge base. The modeling engine uses the feedback to update models in the adaptive knowledge base. The modeling engine supports various application specific modules.

Claims

1. A system for electronic communication management comprising:

a contact center configured to send and receive communications using dissimilar communications channels;

a modeling engine configured to analyze a communication received by the contact center on a communications channel and determine an intent of the received communication, wherein the modeling engine is configured to automatically retrieve a model based on the intent of the received communication;

an automatic response module supported by the modeling engine that generates a predicted response to the received communication using the retrieved model;

an agent that composes an actual response to the received communication based on the intent of the received communication;

an adaptive knowledge base configured to store the models used by the modeling engine; and

a feedback module configured to compare the actual response to the received communication with the predicted response to the received communication and provide feedback to the modeling engine, which uses the feedback to continuously update the models in the adaptive knowledge base, so as to improve subsequent predicted responses to received communications.

2. The system of claim 1, wherein the contact center is configured to send and receive communications via text-based communication channels.

3. The system of claim 1, wherein the contact center is configured to send and receive communications via voice-based communication channels.

4. The system of claim 1, wherein the contact center is configured to receive text communications comprising natural language.

5. The system of claim 4, wherein the modeling engine comprises a natural language processor configured to analyze the text communications to identify concepts.

6. The system of claim 5, wherein the natural language processor performs a morphological analysis of the text communications.

7. The system of claim 5, wherein the natural language processor performs a semantic analysis of the text communications.

8. The system of claim 5, wherein the natural language processor comprises a lexical knowledge base.

9. The system of claim 1, wherein the contact center is configured to convert received communications into a universal data model format.

10. The system of claim 1, further comprising an audit module configured to monitor responses generated by agents for quality.

11. The system of claim 10, wherein the audit module is configured to generate an audit result that is fed back to the modeling engine.

12. The system of claim 1, wherein each of the models in the adaptive knowledge base comprises an accuracy gauge configured to be updated by feedback.

13. The system of claim 12, wherein the adaptive knowledge base comprises models for active concepts and models for inactive concepts.

14. The system of claim 13, wherein the models for active concepts become inactive when they have a sufficiently low accuracy rating.

15. The system of claim 13, wherein the models for inactive concepts become active when they have a sufficiently high accuracy rating.

16. The system of claim 1, wherein the models in the adaptive knowledge base are organized into categories and the categories are associated with branches.

17. The system of claim 16, wherein the modeling engine is configured to modify the branches in the adaptive knowledge base using the feedback from the feedback module.

18. The system of claim 16, wherein hierarchies of the branches in the adaptive knowledge base comprise manually created hierarchies.

19. The system of claim 16, wherein hierarchies of the branches in the adaptive knowledge base comprise automatically created hierarchies.

20. The system of claim 16, wherein the branches in the adaptive knowledge base have associated rules.

21. The system of claim 1, wherein the modeling engine comprises a statistical modeler configured to create the models and perform relationship algebra using the models.

22. The system of claim 1, wherein the modeling engine is configured to support an application specific module.

23. The system of claim 22, wherein the application specific module is an automatic response module.

24. The system of claim 22, wherein the application specific module is an expertise routing module.

25. The system of claim 22, wherein the application specific module is an automatic task prioritization module.

26. The system of claim 22, wherein the application specific module is a content filter module configured to filter content of agent-generated responses.

27. The system of claim 22, wherein the application specific module is a business process automation module.

28. The system of claim 22, wherein the application specific module is a workflow application.

29. The system of claim 22, wherein the application specific module is a Frequently Asked Questions module.

30. The system of claim 22, wherein the application specific module is configured to generally classify the received communications according to content.

31. The system of claim 2, further comprising a digital signal processing module configured to process received voice communications.

32. The system of claim 31, wherein the digital signal processing module is further configured to categorize the received voice communications according to acoustical content of the received voice communications.

33. The system of claim 1, wherein the feedback module is further configured to support multiple feedbacks to a single received communication.

34. The system of claim 1, wherein the received communications comprise documents.

35. The system of claim 34, wherein a statistical matching value between the documents and the models is evaluated by a calculated statistical likelihood value.

36. A method for managing electronic communications in a computer network, the method comprising:

receiving a communication from one of a plurality of dissimilar communications channels over the computer network;

analyzing the communication at a computer coupled to the computer network to determine an intent of the communication, and automatically retrieving a model based on the intent of the communication;

generating a predicted response to the communication using the retrieved model;

generating an actual response to the communication based on the intent of the communication; and

comparing the actual response to the predicted response to provide feedback, which is used to update the model, so as to improve subsequent predicted responses to communications received over the computer network.

37. The method of claim 36, further comprising routing the communication based on semantical content of the communication.

38. The method of claim 36, wherein the communication comprises a communication expressed in natural language.

39. The method of claim 36, wherein generating a predicted response to the communication comprises comparing the communication to a model.

40. The method of claim 36, wherein generating an actual response is performed by an automatic response module.

41. The method of claim 36, wherein generating an actual response is performed by an agent.

42. The method of claim 36, wherein the communication comprises a text communication comprising natural language.

43. The method of claim 42, wherein analyzing the communication comprises morphological analysis or semantic analysis.

44. The method of claim 36, wherein generating a predicted response to the communication comprises comparing the communication to a set of models that correspond to a category related to the intent of the communication.

45. The method of claim 36, wherein comparing the actual response and the predicted response generates feedback that is used to modify a model.

46. The method of claim 45, wherein if the actual response is similar to the predicted response, the generated feedback is positive, and if the actual response is different from the predicted response, the generated feedback is negative.

47. The method of claim 36, wherein the communication comprises a voice communication expressed in natural language.

48. The method of claim 47, wherein analyzing the communication comprises digital signal processing of the voice communication.

49. The method of claim 48, wherein generating the predicted response to the communication comprises categorizing the voice communication based on acoustical content of the voice communication.

50. A method for processing a relationship event in a computer network, the method comprising:

receiving the relationship event over the computer network;

analyzing the relationship event at a computing device coupled to the computer network to identify concepts in the relationship event;

building an event model of the relationship event using the identified concepts;

mapping the event model to models in a knowledge base to generate category scores; and

routing the relationship event over the computer network for action based on the category scores.

51. A computer-readable medium having a program embodied thereon, the program being executable by a computer to perform a method for electronic communication management, the method comprising:

receiving a communication from one of a plurality of dissimilar communications channels;

analyzing the communication to determine an intent of the communication, and automatically retrieving a model based on the intent of the communication;

generating a predicted response to the communication using the retrieved model;

generating an actual response to the communication based on the intent of the communication; and

comparing the actual response and the predicted response to provide feedback, which is used to update the model, so as to improve subsequent predicted responses to communications.

52. The computer-readable medium of claim 51, wherein comparing the actual response and the predicted response occurs in real time.

53. The computer-readable medium of claim 51, wherein comparing the actual response and the predicted response occurs off-line.

54. A computer-readable medium having a program embodied thereon, the program being executable by a computer to perform a method for processing a relationship event, the method comprising:

receiving the relationship event;

analyzing the relationship event to identify concepts in the relationship event;

building an event model of the relationship event using the concepts;

mapping the event model to models in a knowledge base to generate category scores; and

routing the relationship event over the computer network for action based on the category scores.

55. A system for electronic communication management, comprising:

means for receiving a communication from one of a plurality of dissimilar communications channels;

means for analyzing the communication to determine intent, and automatically retrieving a model based on the intent of the communication;

means for predicting a response to the communication using the retrieved model, and for generating a predicted response;

means for preparing a response to the communication based on the intent of the communication, and for generating an actual response; and

means for comparing the actual response and the predicted response to provide feedback, which is used to update the model, so as to improve subsequent predictions.

56. A system for electronic communication management, comprising:

a contact center configured to send and receive communications via dissimilar communication channels;

a modeling engine configured to analyze a received communication to determine an intent, and further configured to retrieve a model based on the intent;

an adaptive knowledge base configured to store models; and

a feedback module configured to compare a response predicted by the modeling engine in conjunction with the models in the adaptive knowledge base and an actual response to the received communication to generate feedback, the feedback being used to continuously update the models in the adaptive knowledge base such that the system learns from each received communication, in order to improve subsequent predicted responses to received communications.

57. The system of claim 56, wherein the modeling engine gains knowledge from communications on one communication channel and applies the knowledge to communications on another communication channel.

58. A method for computerized analysis of communications using computer-generated adaptive models, comprising:

receiving a communication via one of a plurality of dissimilar communications channels;

analyzing content of the communication on a computer to identify at least one concept of the communication;

creating a model of the communication using the at least one concept;

comparing the model of the communication to a set of adaptive models to generate a predicted response to the communication using the model

generating an actual response to the communication based on the concept of the communication;

comparing the predicted response and the actual response to generate feedback; and

using the feedback to update at least one of the set of adaptive models such that the set of adaptive models learns with each received communication, in order to improve subsequent predicted responses to received communications.

59. The method of claim 58, wherein comparing the predicted response and the actual response occurs in real time.

60. The method of claim 58, wherein using the feedback to modify at least one of the set of adaptive models occurs in real time.

61. The method of claim 58, wherein comparing the predicted response and the actual response occurs while further communications are being received.

62. The method of claim 58, wherein using the feedback to modify at least one of the set of adaptive models occurs while further communications are being received.

63. The method of claim 58, wherein the content of the communication comprises content expressed in a natural language.

64. The method of claim 58, wherein the content of the communication comprises metadata.

65. The method of claim 58, wherein the content of the communication comprises structured information.

66. The method of claim 58, wherein the communication comprises a text communication.

67. The method of claim 58, wherein the communication comprises a voice communication.

68. A system for electronic communication management, comprising:

a contact center configured to send and receive communications via dissimilar communications channels;

an adaptive knowledge base configured to store models;

a modeling engine configured to analyze a received communication to determine an intent, to prepare a model of the communication based on the intent, and to compare the model of the communication with the models stored in the adaptive knowledge base to generate a predicted response; and

a feedback module configured to compare the predicted response with an actual response to the received communication to generate feedback used by the adaptive knowledge base to update at least one model such that the system learns from the received communication, in order to improve subsequent predicted responses to received communications.

69. The system of claim 68, wherein a human agent generates the actual response to the received communication.

70. The system of claim 68, wherein the adaptive knowledge base modifies at least one model in response to each communication received by the contact center such that the system learns from each received communication.

71. The system of claim 68, wherein the modeling engine is further configured to determine a plurality of intents in the received communication.

72. The system of claim 71, wherein the modeling engine is further configured to determine an explicit intent and an implicit intent in the received communication.

73. A method for real-time learning in a computerized communication management system, comprising:

receiving a communication using one of a plurality of dissimilar communications channels;

creating a model of the communication on a computer;

comparing the model of the communication to a set of adaptive models to generate a predicted action in response to the communication;

comparing the predicted action with an actual action in response to the communication to generate feedback; and

continuously updating the set of adaptive models according to the feedback so as to improve subsequent predicted responses to received communications.

74. The method of claim 73, wherein if the predicted action is similar to the actual action, the feedback is positive and an accuracy rating of a model in the set of adaptive models that generated the predicted action is increased.

75. The method of claim 73, wherein if the predicted action differs from the actual action, the feedback is negative and an accuracy rating of a model in the set of adaptive models that generated the predicted action is decreased.

76. The method of claim 73, wherein if the predicted action differs from the actual action and if a model that is similar to the actual action exists in the set of adaptive models, then the feedback is negative for a model in the set of adaptive models that generated the predicted action and the feedback is positive for the model that is similar to the actual action.

77. A method for real-time modeling of communications in a computerized communication management system, comprising:

receiving a communication from one of a plurality of dissimilar communications channels;

creating a model of the communication on a computer;

comparing the model of the communication to a set of adaptive models to determine a category for the communication;

comparing the determined category with an actual category for the communication to generate feedback; and

continuously updating the set of adaptive models according to the feedback, so as to improve subsequent predicted responses to received communications.