Granted Patent B2

US 7,818,176 · App. 11/671,526 · Granted Oct 19, 2010

Abstract

Claims (102)

1. A method for selecting and presenting advertisements in response to processing natural language utterances, comprising:

receiving a natural language utterance containing at least one request at an input device;

recognizing one or more words or phrases in the natural language utterance at a speech recognition engine coupled to the input device, wherein recognizing the words or phrases in the natural language utterance includes:

mapping a stream of phonemes contained in the natural language utterance to one or more syllables that are phonemically represented in an acoustic grammar; and

generating a preliminary interpretation for the natural language utterance from the one or more syllables, wherein the preliminary interpretation generated from the one or more syllables includes the recognized words or phrases;

interpreting the recognized words or phrases at a conversational language processor coupled to the speech recognition engine, wherein interpreting the recognized words or phrases includes establishing a context for the natural language utterance;

selecting an advertisement in the context established for the natural language utterance; and

presenting the selected advertisement via an output device coupled to the conversational language processor.

2. The method of claim 1 , wherein the conversational language processor selects the advertisement based on information related to one or more of the recognized words or phrases, an action associated with the request, a personalized cognitive model derived from an interaction pattern for a specific user, a generalized cognitive model derived from an interaction pattern for a plurality of users, or an environmental model derived from environmental conditions or surroundings associated with the specific user.

3. The method of claim 2 , further comprising:

tracking an interaction pattern with the advertisement presented via the output device; and

updating one or more of the personalized cognitive model, the generalized cognitive model, or the environmental model based on the interaction pattern tracked for the advertisement.

4. The method of claim 3 , wherein updating one or more of the personalized cognitive model, the generalized cognitive model, or the environmental model builds statistical profiles for selecting subsequent advertisements in response to subsequent natural language utterances.

5. The method of claim 4 , wherein the statistical profiles identify affinities between the advertisement presented via the output device and one or more of the recognized words or phrases, the action associated with the request, the personalized cognitive model, the generalized cognitive model, or the environmental model.

6. The method of claim 3 , further comprising:

building long-term shared knowledge and short-term shared knowledge in response to updating one or more of the personalized cognitive model, the generalized cognitive model, or the environmental model;

receiving a subsequent natural language utterance at the input device; and

interpreting the subsequent natural language utterance at the conversational language processor using the long-term shared knowledge and the short-term shared knowledge.

7. The method of claim 3 , wherein the interaction pattern tracked for the advertisement includes an action performed in response to a subsequent request that identifies the advertisement.

8. The method of claim 7 , wherein the action includes executing a task or retrieving information based on the subsequent request that identifies the advertisement.

9. The method of claim 1 , wherein the conversational language processor selects the advertisement to resolve the request in response to determining that the natural language utterance includes incomplete or ambiguous information.

10. The method of claim 1 , wherein the speech recognition engine recognizes the one or more words or phrases in the natural language utterance using a plurality of entries in one or more dictionary and phrase tables that are dynamically updated based on a history of a current dialog and one or more prior dialogs.

11. The method of claim 10 , wherein the plurality of entries in the one or more dictionary and phrase tables are further dynamically updated based on one or more dynamic fuzzy set possibilities or prior probabilities derived from the history of the current dialog and the prior dialogs.

12. The method of claim 1 , further comprising determining that the conversational language processor incorrectly interpreted the words or phrases in response to an adaptive misrecognition engine detecting a predetermined event, wherein the conversational language processor reinterprets the words or phrases in response to the predetermined event.

13. The method of claim 12 , wherein the predetermined event includes receiving one or more of a subsequent input proximate in time to the natural language utterance, a subsequent input that stops the conversational language processor from processing the request contained in the natural language utterance, a subsequent input that overrides the request in a time shorter than an expected time to process the request, or a subsequent input that repeats the natural language utterance.

14. A method for selecting and presenting advertisements in response to processing natural language utterances, comprising:

receiving a natural language utterance containing at least one request at an input device;

recognizing one or more words or phrases in the natural language utterance at a speech recognition engine coupled to the input device;

interpreting the recognized words or phrases at a conversational language processor coupled to the speech recognition engine, wherein interpreting the recognized words or phrases includes establishing a context for the natural language utterance;

selecting an advertisement in the context established for the natural language utterance;

presenting the selected advertisement via an output device coupled to the conversational language processor; and

determining that the conversational language processor incorrectly interpreted the words or phrases in response to an adaptive misrecognition engine detecting a predetermined event, wherein the conversational language processor reinterprets the words or phrases in response to the predetermined event.

15. The method of claim 14 , wherein the predetermined event includes receiving one or more of a subsequent input proximate in time to the natural language utterance, a subsequent input that stops the conversational language processor from processing the request contained in the natural language utterance, a subsequent input that overrides the request in a time shorter than an expected time to process the request, or a subsequent input that repeats the natural language utterance.

16. The method of claim 14 , wherein recognizing the words or phrases in the natural language utterance includes:

mapping a stream of phonemes contained in the natural language utterance to one or more syllables that are phonemically represented in an acoustic grammar; and

generating a preliminary interpretation for the natural language utterance from the one or more syllables, wherein the preliminary interpretation generated from the one or more syllables includes the recognized words or phrases.

17. The method of claim 14 , wherein the conversational language processor selects the advertisement to resolve the request in response to determining that the natural language utterance includes incomplete or ambiguous information.

18. The method of claim 14 , wherein the conversational language processor selects the advertisement based on information related to one or more of the recognized words or phrases, an action associated with the request, a personalized cognitive model derived from an interaction pattern for a specific user, a generalized cognitive model derived from an interaction pattern for a plurality of users, or an environmental model derived from environmental conditions or surroundings associated with the specific user.

19. The method of claim 18 , further comprising:

tracking an interaction pattern with the advertisement presented via the output device; and

updating one or more of the personalized cognitive model, the generalized cognitive model, or the environmental model based on the interaction pattern tracked for the advertisement.

20. The method of claim 19 , wherein updating one or more of the personalized cognitive model, the generalized cognitive model, or the environmental model builds statistical profiles for selecting subsequent advertisements in response to subsequent natural language utterances.

21. The method of claim 20 , wherein the statistical profiles identify affinities between the advertisement presented via the output device and one or more of the recognized words or phrases, the action associated with the request, the personalized cognitive model, the generalized cognitive model, or the environmental model.

22. The method of claim 19 , further comprising:

building long-term shared knowledge and short-term shared knowledge in response to updating one or more of the personalized cognitive model, the generalized cognitive model, or the environmental model;

receiving a subsequent natural language utterance at the input device; and

interpreting the subsequent natural language utterance at the conversational language processor using the long-term shared knowledge and the short-term shared knowledge.

23. The method of claim 19 , wherein the interaction pattern tracked for the advertisement includes an action performed in response to a subsequent request that identifies the advertisement.

24. The method of claim 23 , wherein the action includes executing a task or retrieving information based on the subsequent request that identifies the advertisement.

25. The method of claim 14 , wherein the speech recognition engine recognizes the one or more words or phrases in the natural language utterance using a plurality of entries in one or more dictionary and phrase tables that are dynamically updated based on a history of a current dialog and one or more prior dialogs.

26. The method of claim 25 , wherein the plurality of entries in the one or more dictionary and phrase tables are further dynamically updated based on one or more dynamic fuzzy set possibilities or prior probabilities derived from the history of the current dialog and the prior dialogs.

27. A system for selecting and presenting advertisements in response to processing natural language utterances, comprising:

an input device that receives a natural language utterance containing at least one request at an input device;

a speech recognition engine coupled to the input device, wherein the speech recognition engine recognizes one or more words or phrases in the natural language utterance, wherein to recognize the words or phrases in the natural language utterance, the speech recognition engine is configured to:

map a stream of phonemes contained in the natural language utterance to one or more syllables that are phonemically represented in an acoustic grammar; and

generate a preliminary interpretation for the natural language utterance from the one or more syllables, wherein the preliminary interpretation generated from the one or more syllables includes the recognized words or phrases;

a conversational language processor coupled to the speech recognition engine, wherein the conversational language processor is configured to:

interpret the recognized words or phrases, wherein interpreting the recognized words or phrases includes establishing a context for the natural language utterance;

select an advertisement in the context established for the natural language utterance; and

present the selected advertisement via an output device.

28. The system of claim 27 , wherein the conversational language processor selects the advertisement based on information related to one or more of the recognized words or phrases, an action associated with the request, a personalized cognitive model derived from an interaction pattern for a specific user, a generalized cognitive model derived from an interaction pattern for a plurality of users, or an environmental model derived from environmental conditions or surroundings associated with the specific user.

29. The system of claim 28 wherein the conversational language processor is further configured to:

track an interaction pattern with the advertisement presented via the output device; and

update one or more of the personalized cognitive model, the generalized cognitive model, or the environmental model based on the interaction pattern tracked for the advertisement.

30. The system of claim 29 , wherein the conversational language processor updates one or more of the personalized cognitive model, the generalized cognitive model, or the environmental model to build statistical profiles for selecting subsequent advertisements in response to subsequent natural language utterances.

31. The system of claim 30 , wherein the statistical profiles identify affinities between the advertisement presented via the output device and one or more of the recognized words or phrases, the action associated with the request, the personalized cognitive model, the generalized cognitive model, or the environmental model.

32. The system of claim 29 , wherein the conversational language processor is further configured to:

build long-term shared knowledge and short-term shared knowledge in response to updating one or more of the personalized cognitive model, the generalized cognitive model, or the environmental model; and

interpret a subsequent natural language utterance received at the input device language processor using the long-term shared knowledge and the short-term shared knowledge.

33. The system of claim 29 , wherein the interaction pattern tracked for the advertisement includes an action performed in response to a subsequent request that identifies the advertisement.

34. The system of claim 33 , wherein the action includes executing a task or retrieving information based on the subsequent request that identifies the advertisement.

35. The system of claim 27 , wherein the conversational language processor selects the advertisement to resolve the request in response to determining that the natural language utterance includes incomplete or ambiguous information.

36. The system of claim 27 wherein the speech recognition engine recognizes the one or more words or phrases in the natural language utterance using a plurality of entries in one or more dictionary and phrase tables that are dynamically updated based on a history of a current dialog and one or more prior dialogs.

37. The system of claim 36 , wherein the plurality of entries in the one or more dictionary and phrase tables are further dynamically updated based on one or more dynamic fuzzy set possibilities or prior probabilities derived from the history of the current dialog and the prior dialogs.

38. The system of claim 27 , further comprising an adaptive misrecognition engine configured to determine that the conversational language incorrectly interpreted the words or phrases in response to detecting a predetermined event, wherein the conversational language processor reinterprets the words or phrases in response to the predetermined event.

39. The system of claim 38 , wherein the predetermined event includes receiving one or more of a subsequent input proximate in time to the natural language utterance, a subsequent input that stops the conversational language processor from processing the request contained in the, natural language utterance, a subsequent input that overrides the request in a time shorter than an expected time to process the request, or a subsequent input that repeats the natural language utterance.

40. A system for selecting and presenting advertisements in response to processing natural language utterances, comprising:

an input device that receives a natural language utterance containing at least one request at an input device;

a speech recognition engine coupled to the input device, wherein the speech recognition engine recognizes one or more words or phrases in the natural language utterance;

a conversational language processor coupled to the speech recognition engine, wherein the conversational language processor is configured to:

interpret the recognized words or phrases, wherein interpreting the recognized words or phrases includes establishing a context for the natural language utterance;

select an advertisement in the context established for the natural language utterance; and

present the selected advertisement via an output device; and

an adaptive misrecognition engine configured to determine that the conversational language incorrectly interpreted the words or phrases in response to detecting a predetermined event, wherein the conversational language processor reinterprets the words or phrases in response to the predetermined event.

41. The system of claim 40 , wherein the predetermined event includes receiving one or more of a subsequent input proximate in time to the natural language utterance, a subsequent input that stops the conversational language processor from processing the request contained in the natural language utterance, a subsequent input that overrides the request in a time shorter than an expected time to process the request, or a subsequent input that repeats the natural language utterance.

42. The system of claim 40 , wherein to recognize the words or phrases in the natural language utterance, the speech recognition engine is configured to:

map a stream of phonemes contained in the natural language utterance to one or more syllables that are phonemically represented in an acoustic grammar; and

generate a preliminary interpretation for the natural language utterance from the one or more syllables, wherein the preliminary interpretation generated from the one or more syllables includes the recognized words or phrases.

43. The system of claim 40 , wherein the conversational language processor selects the advertisement to resolve the request in response to determining that the natural language utterance includes incomplete or ambiguous information.

44. The system of claim 40 , wherein the conversational language processor selects the advertisement based on information related to one or more of the recognized words or phrases, an action associated with the request, a personalized cognitive model derived from an interaction pattern for a specific user, a generalized cognitive model derived from an interaction pattern for a plurality of users, or an environmental model derived from environmental conditions or surroundings associated with the specific user.

45. The system of claim 44 , wherein the conversational language processor is further configured to:

track an interaction pattern with the advertisement presented via the output device; and

update one or more of the personalized cognitive model, the generalized cognitive model, or the environmental model based on the interaction pattern tracked for the advertisement.

46. The system of claim 45 , wherein the conversational language processor updates one or more of the personalized cognitive model, the generalized cognitive model, or the environmental model to build statistical profiles for selecting subsequent advertisements in response to subsequent natural language utterances.

47. The system of claim 46 , wherein the statistical profiles identify affinities between the advertisement presented via the output device and one or more of the recognized words or phrases, the action associated with the request, the personalized cognitive model, the generalized cognitive model, or the environmental model.

48. The system of claim 45 , wherein the conversational language processor is further configured to:

build long-term shared knowledge and short-term shared knowledge in response to updating one or more of the personalized cognitive model, the generalized cognitive model, or the environmental model; and

interpret a subsequent natural language utterance received at the input device language processor using the long-term shared knowledge and the short-term shared knowledge.

49. The system of claim 45 , wherein the interaction pattern tracked for the advertisement includes an action performed in response to a subsequent request that identifies the advertisement.

50. The system of claim 49 , wherein the action includes executing a task or retrieving information based on the subsequent request that identifies the advertisement.

51. The system of claim 40 , wherein the speech recognition engine recognizes the one or more words or phrases in the natural language utterance using a plurality of entries in one or more dictionary and phrase tables that are dynamically updated based on a history of a current dialog and one or more prior dialogs.

52. The system of claim 51 wherein the plurality of entries in the one or more dictionary and phrase tables are further dynamically updated based on one or more dynamic fuzzy set possibilities or prior probabilities derived from the history of the current dialog and the prior dialogs.

Assignments (10)