Recognition of proper nouns using native-language pronunciation

• US 8,285,537 B2
• Filed: 01/31/2003
• Issued: 10/09/2012
• Est. Priority Date: 01/31/2003
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Comverse Incorporated (Mavenir Group Holdings Ltd.)
• Inventors: Tanner, Marc D., Panttaja, Erin M.
• Primary Examiner(s): Saint Cyr, Leonard
• Application Number:
  US10/355,148
• Publication Number:
  US 20040153306A1
• Time in Patent Office: 3,539 Days
• Field of Search: None
• US Class Current: 704/4
• CPC Class Codes: G06F 40/232 Orthographic correction, e.g. spell checking or vowelisation
• CPC Class Codes: G06F 40/242 Dictionaries
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams

Abstract

Recognition of proper nouns by an automated speech recognition system is improved by augmenting the pronunciation of each proper noun or name in the natural language of the speech recognition system with at least one “native” pronunciation in another natural language. To maximize recognition, preferably the pronunciations are predicted based on information not available to the speech recognition system. Prediction of pronunciation may be based on a location derived from a telephone number or postal address associated with the name and the language or dialect spoken in the country or region of that location. The “native” pronunciation(s) may be added to a dictionary of the speech recognition system or directly to the grammar used for recognizing speech.

Claims

1. A method of improving automated recognition of proper nouns spoken by users, comprising:

predicting at least one native pronunciation, in at least one natural language, of a proper noun stored as text, based on information associated with the proper noun but not derived from the text of the proper noun; and

augmenting a speech recognizer, that is not configured to recognize words in the at least one natural language, using the at least one native pronunciation of the proper noun.

2. A method as recited in claim 1, wherein said predicting uses data available to an application other than the speech recognizer.

3. A method as recited in claim 2, wherein said predicting is based on a country associated with the proper noun.

4. A method as recited in claim 3, wherein said predicting also uses spelling of the proper noun.

5. A method as recited in claim 3, wherein the country is derived from a postal address associated with the proper noun.

6. A method as recited in claim 5, wherein said predicting also uses spelling of the proper noun.

7. A method as recited in claim 3, wherein the proper noun is a name and the country is derived from a telephone number associated with the name.

8. A method as recited in claim 7, wherein said predicting also uses spelling of the proper noun.

9. A method as recited in claim 2 wherein said predicting is based on at least one of information about a user and a response by the user to an application supplying speech input from the user to the speech recognizer.

10. A method as recited in claim 2,

wherein the speech recognizer is able to recognize words in a first language, and

wherein the proper noun is a name and said predicting includes comparing the name with a list of names associated with at least one language different from the first language.

11. A method as recited in claim 10, wherein said augmenting obtains the at least one native pronunciation of the proper noun from a list of phonetic spellings at least partially created manually.

12. A method as recited in claim 11, further comprising generating the at least one native pronunciation for at least one geographical area using at least one automated tool for text-to-speech conversion.

13. A method as recited in claim 2, wherein said augmenting includes adding at least one native pronunciation of the proper noun to a dictionary used by the speech recognizer.

14. A method as recited in claim 13,

further comprising storing at least one native pronunciation of the proper noun represented in a language independent phonetic alphabet;

and

wherein said augmenting includes converting a representation of the at least one native pronunciation of the proper noun in the language independent phonetic alphabet into a language-specific phonetic alphabet used by the speech recognizer prior to said adding of the at least one native pronunciation to the dictionary.

15. A method as recited in claim 13,

further comprising storing at least one native pronunciation of the proper noun represented in a first phonetic alphabet specifically for a first language;

and

wherein said augmenting includes converting a representation of the at least one native pronunciation of the proper noun in the first phonetic alphabet into a second phonetic alphabet specifically for a second language used by the speech recognizer prior to said adding of the at least one native pronunciation to the dictionary.

16. A method as recited in claim 13, wherein said augmenting uses a standard phonetic alphabet to add the at least one native pronunciation of the proper noun to the dictionary.

17. A method as recited in claim 13, wherein said augmenting uses an approximation of the at least one native pronunciation of the proper noun in a phonetic alphabet of the recognizer.

18. A method as recited in claim 13, wherein said augmenting adds a plurality of different native pronunciations of the proper noun to the dictionary.

19. A method as recited in claim 18, wherein the dictionary ordinarily has one record with one pronunciation per proper noun and said augmenting adds the different native pronunciations in separate records for the proper noun.

20. A method as recited in claim 2, wherein said method further comprises converting the at least one native pronunciation of the proper noun into a format of a grammar used by the speech recognizer.

21. A method as recited in claim 20, wherein the speech recognizer is able to recognize words in a first language, while the native pronunciation of the proper noun is for a second language different than the first language.

22. A method as recited in claim 21, wherein said converting includes approximating the native pronunciation of the proper noun in the second language in text to which pronunciation rules of the first language will be applied by the speech recognizer.

23. A method as recited in claim 2, wherein said augmenting adds a plurality of different native pronunciations of the proper noun to a grammar used by the speech recognizer.

24. A method as recited in claim 2, wherein said augmenting includes approximating the native pronunciation of the proper noun in the second language in text to which pronunciation rules of the first language will be applied by the speech recognizer.

25. A method as recited in claim 2,

further comprising storing at least one native pronunciation of the proper noun represented in a language independent phonetic alphabet;

and

wherein said augmenting includes converting a representation of the at least one native pronunciation of the proper noun in the language independent phonetic alphabet into a language-specific phonetic alphabet used by the speech recognizer prior to said augmenting.

26. A method as recited in claim 2,

further comprising storing at least one native pronunciation of the proper noun represented in a first phonetic alphabet specifically for a first language;

and

wherein said augmenting includes

converting a representation of the at least one native pronunciation of the proper noun in the first phonetic alphabet into a second phonetic alphabet specifically for a second language used by the speech recognizer; and

adding the at least one native pronunciation as represented in the second phonetic alphabet for use by the speech recognizer.

27. A method as recited in claim 2,

wherein the speech recognizer is able to recognize words in a first language,

wherein said predicting identifies multiple native pronunciations for the proper noun, including the at least one native pronunciation in at least one second language different from the first language, and

wherein said augmenting adds the at least one native pronunciation in the at least one second language.

28. A method as recited in claim 27, wherein each native pronunciation of the proper noun is added as an alternative pronunciation of the proper noun.

29. A method as recited in claim 27,

wherein each native pronunciation of the proper noun is added as a native pronunciation of a different proper noun, and

wherein said method further comprises supplying an identified proper noun to an application that determines the identified proper noun corresponds to the proper noun stored as text for which the at least one native pronunciation was predicted.

30. A method as recited in claim 29, wherein said augmenting includes approximating each native pronunciation of the proper noun in the second language in text to which pronunciation rules of the first language will be applied by the speech recognizer.

31. A method as recited in claim 2, further comprising synthesizing speech by the application using the native pronunciation corresponding to the identified proper noun.

32. A method as recited in claim 1, wherein the information associated with the proper noun is not based on characteristics of the users.

33. A method of improving automated recognition of proper nouns spoken by users, comprising:

predicting pronunciation of a proper noun in text, based on information associated with the proper noun but not obtained from the text containing the proper noun; and

generating a text string that, when pronounced as a word in a first natural language, sounds similar to the proper noun when pronounced in a second natural language selected based on said predicting.

34. A method as recited in claim 33, wherein said predicting is based on a country associated with the proper noun.

35. A method as recited in claim 33, wherein the country is derived from a postal address associated with the proper noun.

36. A method as recited in claim 35, wherein the proper noun is a name and the country is derived from a telephone number associated with the name.

37. A method as recited in claim 36, wherein said predicting identifies multiple pronunciations for the proper noun and said generating uses at least some of the multiple pronunciations.

38. A method as recited in claim 37, wherein said predicting uses data obtained from at least one automated tool for text-to-speech conversion that provides the at least one native pronunciation for at least one geographical area.

39. A method as recited in claim 33, wherein the information associated with the proper noun is not based on characteristics of the users.

40. At least one computer readable medium storing at least one program for controlling a computer system to automatically recognize proper nouns spoken by users according to a method comprising:

predicting pronunciation of a proper noun in text, based on information associated with the proper noun but not obtained from the text containing the proper noun; and

generating a text string that, when pronounced as a word in a first natural language, sounds similar to the proper noun when pronounced in a second natural language selected based on said predicting.

41. At least one computer readable medium as recited in claim 40, wherein said predicting is based on a country associated with the proper noun.

42. At least one computer readable medium as recited in claim 40, wherein the country is derived from a postal address associated with the proper noun.

43. At least one computer readable medium as recited in claim 42, wherein the proper noun is a name and the country is derived from a telephone number associated with the name.

44. At least one computer readable medium as recited in claim 43, wherein said predicting identifies multiple pronunciations for the proper noun and said generating uses at least some of the multiple pronunciations.

45. At least one computer readable medium as recited in claim 44, wherein said predicting uses data obtained from at least one automated tool for text-to-speech conversion that provides the at least one native pronunciation for at least one geographical area.

46. At least one computer readable medium as recited in claim 40, further comprising augmenting a speech recognizer, configured to recognize words in the first natural language, but not the second natural language, using the text string produced by said generating.

47. At least one computer readable medium as recited in claim 46, wherein said predicting is based on a country associated with the proper noun.

48. At least one computer readable medium as recited in claim 47, wherein the country is derived from a postal address associated with the proper noun.

49. At least one computer readable medium as recited in claim 47, wherein said predicting also uses spelling of the proper noun.

50. At least one computer readable medium as recited in claim 47, wherein the proper noun is a name and the country is derived from a telephone number associated with the name.

51. At least one computer readable medium as recited in claim 46 wherein said predicting is based on at least one of information about a user and a response by the user to an application supplying speech input from the user to the speech recognizer.

52. At least one computer readable medium as recited in claim 46, wherein the proper noun is a name and said predicting includes comparing the name with a list of names associated with at least one language different from the first language.

53. At least one computer readable medium as recited in claim 46, wherein said augmenting includes adding the text string to a dictionary used by the speech recognizer.

54. At least one computer readable medium as recited in claim 53, wherein said augmenting uses a standard phonetic alphabet in the text string.

55. At least one computer readable medium as recited in claim 53, wherein said augmenting uses a phonetic alphabet of the recognizer in the text string.

56. At least one computer readable medium as recited in claim 53, wherein the dictionary ordinarily has one record with one pronunciation per proper noun and said augmenting adds a plurality of text strings representing different native pronunciations in separate records for the proper noun.

57. At least one computer readable medium as recited in claim 46, wherein said augmenting adds a plurality of text strings representing different native pronunciations of the proper noun.

58. At least one computer readable medium as recited in claim 57, wherein each native pronunciation of the proper noun is added as an alternative pronunciation of the proper noun.

59. At least one computer readable medium as recited in claim 57,

wherein each native pronunciation of the proper noun is added as a native pronunciation of a different proper noun, and

wherein said method further comprises supplying an identified proper noun to an application that determines the identified proper noun corresponds to the proper noun stored as text for which the at least one native pronunciation was predicted.

60. At least one computer readable medium as recited in claim 46, further comprising converting a native pronunciation of the proper noun into the text string in a format of a grammar used by the speech recognizer.

61. At least one computer readable medium as recited in claim 40, wherein the information associated with the proper noun is not based on characteristics of the users.