Natural language dialogue method and natural language dialogue system

• US 9,767,796 B2
• Filed: 12/31/2013
• Issued: 09/19/2017
• Est. Priority Date: 12/31/2012
• Status: Active Grant

• Current Assignee: VIA Technologies Incorporated (VIA Technologies)
• Original Assignee: VIA Technologies Incorporated (VIA Technologies)
• Inventors: Zhang, Guo-Feng
• Primary Examiner(s): SHIN, SEONG-AH A
• Application Number:
  US14/144,578
• Publication Number:
  US 20140188478A1
• Time in Patent Office: 1,358 Days
• Field of Search: 704 8, 704 9, 704 10, 704235, 704254, 707706, 707732
• CPC Class Codes: G06F 16/903 Querying for retrieval from the web G06F16/953
• CPC Class Codes: G10L 15/1822 Parsing for meaning understanding
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 2015/088 Word spotting

Abstract

A natural language dialog method and a natural language dialog system are provided. In the method, a first speech input is received and parsed to generate at least one keyword included in the first speech input, so that a candidate list including at least one report answer is obtained. According to a properties database, one report answer is selected from the candidate list, and a first speech response is output according to the report answer. Other speech inputs are received, and a user'"'"'s preference data is captured from the speech inputs. The user'"'"'s preference data is stored in the properties database.

Claims

1. A natural language dialogue system comprising:

a properties database storing a plurality of records, wherein each of the records comprises a title field, the title field comprises at least one sub-field, each of the at least one sub-field comprises an indication field and a value field, the indication field stores indication data, and the value field stores value data;

a speech sampling module receiving a first speech input, wherein the first speech input is a sentence; and

a natural language comprehension system coupled to the properties database and the speech sampling module, the natural language comprehension system capturing a noun of the sentence as a first keyword, parsing the first keyword included in the first speech input and executing a full-text search in a structured database according to the first keyword to obtain at least one report answer to form a candidate list, the natural language comprehension system selecting the at least one report answer from the candidate list according to the properties database and outputting a first speech response that is a speech converted by the at least one report answer, each of the report answers has a priority, the priorities of the report answers of the candidate list are determined according to usages of the report answers, the first report answer is one of the report answers with the highest priority, the speech sampling module receiving a plurality of speech inputs other than the first speech input, the natural language comprehension system capturing a user'"'"'s preference data from the speech inputs other than the first speech inputs and storing the user'"'"'s preference data into the properties database,

when the speech inputs other than the first speech input present the at least one report answer is incorrect, the at least one report answer is removed from the candidate list,

wherein the first speech response is a sentence outputted through a speech output interface,

wherein the full-text search compares the value data of each of the records and the first keyword to determine whether each of the records is matched and selects the report answers by the matched records.

2. The natural language dialogue system as recited in claim 1, wherein the natural language comprehension system arranges the at least one report answer in the candidate list according to the priority.

3. The natural language dialogue system as recited in claim 2, wherein the natural language comprehension system further determines the priority of the at least one report answer according to the user'"'"'s preference data stored in the properties database.

4. The natural language dialogue system as recited in claim 2, wherein the properties database further stores a user'"'"'s habit, and the natural language comprehension system determines the priority of the at least one report answer according to the user'"'"'s habit.

5. The natural language dialogue system as recited in claim 2, wherein the properties database further stores a public usage habit, and the natural language comprehension system determines the priority of the at least one report answer according to the public usage habit.

6. The natural language dialogue system as recited in claim 1, further comprising:

a speech synthesis database coupled to the natural language comprehension system, wherein the natural language comprehension system searches a first speech corresponding to the first keyword of the first speech input from the speech synthesis database, and the searched first speech acts as the first speech response corresponding to the at least one report answer.

7. The natural language dialogue system as recited in claim 1, the natural language comprehension system parsing a second speech input to obtain a second keyword, wherein the second speech input is received by the speech sampling module before the first speech response is output, and the natural language comprehension system selects one of the at least one report answer matching the first keyword and the second keyword.

8. The natural language dialogue system as recited in claim 1, the natural language dialogue system being suitable for a mobile communication apparatus.

9. A natural language dialogue method comprising:

receiving a first speech input, wherein the first speech input is a sentence;

capturing a noun of the sentence as a first keyword from the first speech input;

parsing the first keyword included in the first speech input, and executing a full-text search in a structured database according to the first keyword to obtain at least one report answer to form a candidate list;

selecting the at least one report answer from the candidate list according to a properties database, wherein each of the report answers has a priority, the priorities of the report answers of the candidate list are determined according to usages of the report answers, the first report answer is one of the report answers with the highest priority, wherein the properties database stores a plurality of records, each of the records comprises a title field, the title field comprises at least one sub-field, each of the at least one sub-field comprises an indication field and a value field, the indication field stores indication data, and the value field stores value data;

outputting a first speech response that is converted by the at least one report answer, wherein the first speech response is a sentence outputted through a speech output interface;

receiving a plurality of speech inputs other than the first speech input;

capturing a user'"'"'s preference data from the speech inputs other than the first speech input; and

storing the user'"'"'s preference data into the properties database,

wherein when the speech inputs other than the first speech input present the at least one report answer is incorrect, the at least one report answer is removed from the candidate list,

wherein the full-text search compares the value data of each of the records and the first keyword to determine whether each of the records is matched and selects the report answers by the matched records.

10. The natural language dialogue method as recited in claim 9, further comprising:

arranging the at least one report answer in the candidate list according to the priority.

11. The natural language dialogue method as recited in claim 10, further comprising:

determining the priority of the at least one report answer according to the user'"'"'s preference data stored in the properties database.

12. The natural language dialogue method as recited in claim 10, further comprising:

storing a user'"'"'s habit into the properties database; and

determining the priority of the at least one report answer according to the user'"'"'s habit.

13. The natural language dialogue method as recited in claim 10, further comprising:

storing a public usage habit into the properties database;

determining the priority of the at least one report answer according to the public usage habit.

14. The natural language dialogue method as recited in claim 10, wherein the step of selecting the at least one report answer comprises:

in a speech synthesis database, searching a first speech corresponding to the first keyword included in the first speech input; and

selecting one of the at least one report answer conforming to the first speech.

15. The natural language dialogue method as recited in claim 10, wherein the step of selecting the at least one report answer further comprises:

parsing a second speech input to obtain at least one second keyword, wherein the second speech input is input before the first speech input; and

selecting one of the at least one report answer matching the first keyword and the at least one second keyword.

16. The natural language dialogue method as recited in claim 9, the natural language dialogue method being suitable for a mobile communication apparatus.

17. A language dialogue system comprising:

a properties database for storing a user'"'"'s preference data and a plurality of records, wherein each of the records comprises a title field, the title field comprises at least one sub-field, each of the at least one sub-field comprises an indication field and a value field, the indication field stores indication data, and the value field stores value data; and

a language comprehension system capturing a noun of the sentence as a first keyword from a first speech input comprising a sentence, parsing a first keyword included in the first speech input and executing a full-text search in a structured database according to the first keyword to obtain at least one report answer to form a candidate list, the language comprehension system selects one of the at least one first report answer from the candidate list according to the user'"'"'s preference data stored in the properties database and outputs a first speech response that is converted by the selected one of the at least one first report answer, each of the report answers has a priority, the priorities of the report answers of the candidate list are determined according to usages of the report answers, the first report answer is one of the report answers with the highest priority, wherein the language comprehension system obtains a first user'"'"'s preference data according to the first speech input and revises the user'"'"'s preference data stored into the properties database according to the first user'"'"'s preference data,

when the speech inputs other than the first speech input present the at least one report answer is incorrect, the at least one report answer is removed from the candidate list,

wherein the first speech response is a sentence outputted through a speech output interface,

wherein the full-text search compares the value data of each of the records and the first keyword to determine whether each of the records is matched and selects the report answers by the matched records.

18. The language dialogue system as recited in claim 17, wherein the language comprehension system determines the priority of the at least one first report answer according to the user'"'"'s preference data.

19. The language dialogue system as recited in claim 18, wherein the user'"'"'s preference data stored into the properties database comprises a correlation between a specific keyword and a user'"'"'s personal preference, a user'"'"'s personal usage habit, a public preference, or a public usage habit.

20. The language dialogue system as recited in claim 17, wherein each of the records further comprises a keyword and a corresponding user'"'"'s personal preference record data or a corresponding user'"'"'s personal dislike record data.

21. The language dialogue system as recited in claim 20, wherein the step of selecting the one of the at least one first report answer from the candidate list according to the user'"'"'s preference data stored in the properties database comprises:

searching one of the records corresponding to the first keyword from the records stored in the properties database; and

selecting the one of the at least one first report answer according to the user'"'"'s personal preference record data included in the corresponding record, the user'"'"'s personal dislike record data included in the corresponding record, or a combination of the user'"'"'s personal preference record data and the user'"'"'s personal dislike record data included in the corresponding record.

22. The language dialogue system as recited in claim 20, wherein the records included in the properties database further comprises a user'"'"'s personal downloaded record corresponding to the keyword, a user'"'"'s personal cited record corresponding to the keyword, a user'"'"'s personal recommendation record corresponding to the keyword, a user'"'"'s personal comment record corresponding to the keyword, or a user'"'"'s personal referral record corresponding to the keyword.

23. The language dialogue system as recited in claim 22, wherein the at least one first report answer is selected according to the user'"'"'s personal downloaded record data included in the records, the user'"'"'s personal cited record data included in the records, the user'"'"'s personal recommendation record data included in the records, the user'"'"'s personal comment record data included in the records, the user'"'"'s personal referral record data included in the records, or at least a combination of two of said user'"'"'s personal record data.

24. The language dialogue system as recited in claim 20, wherein the step of revising the user'"'"'s preference data stored into the properties database according to the first user'"'"'s preference data comprises:

searching one of the records corresponding to the first keyword from the records stored in the properties database; and

in the corresponding record of the first keyword, revising the user'"'"'s personal preference record data or the user'"'"'s personal dislike record data included in the corresponding record according to a preference indication corresponding to the first keyword included in the first user'"'"'s preference data.

25. The language dialogue system as recited in claim 20, further comprising if the one of the records corresponding to the first keyword is not searched from the properties database, a new record is established in the properties database according to the first keyword, and the user'"'"'s personal preference record data corresponding to the new record and the user'"'"'s personal dislike record data corresponding to the new record are revised according to the user'"'"'s preference indication corresponding to the first keyword.

26. The language dialogue system as recited in claim 24, wherein when the records stored in the properties database corresponds to the first keyword, it means that the keyword included in the records stored in the properties database corresponds to the first keyword.

27. The language dialogue system as recited in claim 17, wherein each of the records further comprises a keyword and a corresponding public preference record data or a corresponding public dislike record data.

28. The language dialogue system as recited in claim 27, wherein the step of selecting the one of the at least one first report answer from the candidate list according to the user'"'"'s preference data stored in the properties database comprises:

searching one of the records corresponding to the first keyword according to the keyword of the records stored in the properties database;

selecting the one of the at least one first report answer according to the public preference record data included in the corresponding record, the public dislike record data included in the corresponding record, or a combination of the public preference record data and the public dislike record data included in the corresponding record.

29. The language dialogue system as recited in claim 27, wherein the records included in the properties database further comprises a public downloaded record data corresponding to the keyword, a public cited record data corresponding to the keyword, a public recommendation record data corresponding to the keyword, a public comment record data corresponding to the keyword, or a public referral record data corresponding to the keyword.

30. The language dialogue system as recited in claim 29, wherein the at least one first report answer is selected according to the public downloaded record data included in the records, the public cited record data included in the records, the public recommendation record data included in the records, the public comment record data included in the records, the public referral record data included in the records, or at least a combination of two of said public record data.

31. The language dialogue system as recited in claim 27, wherein the step of revising the user'"'"'s preference data stored into the properties database according to the first user'"'"'s preference data comprises:

searching one of the records corresponding to the first keyword from the records stored in the properties database; and

in the record corresponding to the first keyword, revising the public preference record data or the public dislike record data included in the corresponding record according to a user'"'"'s preference indication corresponding to the first keyword included in the first user'"'"'s preference data.

32. The language dialogue system as recited in claim 31, wherein if the one of the records corresponding to the first keyword is not searched from the properties database, a new record is established in the properties database according to the first keyword, and the public preference record data corresponding to the new record and the public dislike record data corresponding to the new record are revised according to the user'"'"'s preference indication corresponding to the first keyword.

33. The language dialogue system as recited in claim 17, further comprising:

parsing a second speech input to obtain at least one second keyword, wherein the second speech input is input by a user after the first speech response is output; and

selecting a second report answer other than the at least one first report answer from the candidate list according to the second keyword.

34. The language dialogue system as recited in claim 33, further comprising the language comprehension system captures a second user'"'"'s preference data from the second speech input and revises the user'"'"'s preference data stored into the properties database according to the second user'"'"'s preference data.

35. The language dialogue system as recited in claim 17, wherein the step of revising the user'"'"'s preference data stored into the properties database according to the first user'"'"'s preference data comprises:

searching one of the records corresponding to the first keyword from the records stored in the properties database; and

in the record corresponding to the first keyword, revising the stored user'"'"'s preference data corresponding to the first keyword according to a user'"'"'s preference indication corresponding to the first keyword included in the first user'"'"'s preference data.

36. The language dialogue system as recited in claim 35, wherein if the one of the records corresponding to the first keyword is not searched from the properties database, a new record is established in the properties database according to the first keyword, and the user'"'"'s preference record corresponding to the new record is revised according to the user'"'"'s preference indication corresponding to the first keyword.

37. A language dialogue method comprising:

receiving a first speech input, wherein the first speech input is a sentence;

capturing a noun of the sentence as a first keyword from the first speech input;

parsing the first keyword included in the first speech input and executing a full-text search in a structured database according to the first keyword to obtain at least one report answer to form a candidate list;

selecting the at least one first report answer from the candidate list according to a user'"'"'s preference data stored in a properties database, wherein each of the report answers has a priority, the priorities of the report answers of the candidate list are determined according to usages of the report answers, the first report answer is one of the report answers with the highest priority, wherein the properties database stores a plurality of records, each of the records comprises a title field, the title field comprises at least one sub-field, each of the at least one sub-field comprises an indication field and a value field, the indication field stores indication data, and the value field stores value data;

outputting a first speech response that is a speech converted by the at least one first report answer;

obtaining a first user'"'"'s preference data according to the first speech response; and

revising the user'"'"'s preference data stored in the properties database according to the first user'"'"'s preference data,

when the speech inputs other than the first speech input present the at least one report answer is incorrect, the at least one report answer is removed from the candidate list,

wherein the first speech response is a sentence outputted through a speech output interface,

wherein the full-text search compares the value data of each of the records and the first keyword to determine whether each of the records is matched and selects the report answers by the matched records.

38. The language dialogue method as recited in claim 37, further comprising arranging the at least one first report answer according to the priority determining by the user'"'"'s preference data.

39. The language dialogue method as recited in claim 37, wherein the user'"'"'s preference data comprises a correlation between a specific keyword and a user'"'"'s personal preference, a user'"'"'s personal usage habit, a public preference, or a public usage habit.

40. The language dialogue method as recited in claim 37, wherein each of the records further comprises a keyword and a corresponding user'"'"'s personal preference record data or a corresponding user'"'"'s personal dislike record data.

41. The language dialogue method as recited in claim 40, wherein the step of selecting the one of the at least one first report answer from the candidate list according to the user'"'"'s preference data stored in the properties database comprises:

searching one of the records corresponding to the first keyword from the records stored in the properties database; and

selecting the one of the at least one first report answer according to the user'"'"'s personal preference record included in the corresponding record data, the user'"'"'s personal dislike record data included in the corresponding record, or a combination of the user'"'"'s personal preference record data and the user'"'"'s personal dislike record data included in the corresponding record.

42. The language dialogue method as recited in claim 40, wherein the records included in the properties database further comprises a user'"'"'s personal downloaded record data corresponding to the keyword, a user'"'"'s personal cited record data corresponding to the keyword, a user'"'"'s personal recommendation record data corresponding to the keyword, a user'"'"'s personal comment record data corresponding to the keyword, or a user'"'"'s personal referral record data corresponding to the keyword.

43. The language dialogue method as recited in claim 42, wherein the at least one first report answer is selected according to the user'"'"'s personal downloaded record data included in the records, the user'"'"'s personal cited record data included in the records, the user'"'"'s personal recommendation record data included in the records, the user'"'"'s personal comment record data included in the records, the user'"'"'s personal referral record data included in the records, or at least a combination of two of said user'"'"'s personal record data.

44. The language dialogue method as recited in claim 40, wherein the step of revising the user'"'"'s preference data stored into the properties database according to the first user'"'"'s preference data comprises:

searching one of the records corresponding to the first keyword from the records stored in the properties database; and

in the record corresponding to the first keyword, revising the user'"'"'s personal preference data record or the user'"'"'s personal dislike record data included in the corresponding record according to a user'"'"'s preference indication corresponding to the first keyword included in the first user'"'"'s preference data.

45. The language dialogue method as recited in claim 44, further comprising if the one of the records corresponding to the first keyword is not searched from the properties database, a new record is established in the properties database according to the first keyword, and the user'"'"'s personal preference record data corresponding to the new record and the user'"'"'s personal dislike record data corresponding to the new record are revised according to the user'"'"'s preference indication corresponding to the first keyword.

46. The language dialogue method as recited in claim 44, wherein when the records stored in the properties database corresponds to the first keyword, it means that the keyword included in the records stored in the properties database corresponds to the first keyword.

47. The language dialogue method as recited in claim 37, wherein each of the the records further comprises a keyword and a corresponding public preference record data or a corresponding public dislike record data.

48. The language dialogue method as recited in claim 47, wherein the step of selecting the one of the at least one first report answer from the candidate list according to the user'"'"'s preference data stored in the properties database comprises:

searching one of the records corresponding to the first keyword according to the keyword of the records stored in the properties database;

selecting the one of the at least one first report answer according to the public preference record data included in the corresponding record data, the public dislike record data included in the corresponding record data, or a combination of the public preference record data and the public dislike record data included in the corresponding record.

49. The language dialogue method as recited in claim 47, wherein the records included in the properties database further comprises a public downloaded record data corresponding to the keyword, a public cited record data corresponding to the keyword, a public recommendation record data corresponding to the keyword, a public comment record data corresponding to the keyword, or a public referral record data corresponding to the keyword.

50. The language dialogue method as recited in claim 49, wherein the at least one first report answer is selected according to the public downloaded record data included in the records, the public cited record data included in the records, the public recommendation record data included in the records, the public comment record data included in the records, the public referral record data included in the records, or at least a combination of two of said public record data.

51. The language dialogue method as recited in claim 47, wherein the step of revising the user'"'"'s preference data stored into the properties database according to the first user'"'"'s preference data comprises:

searching one of the records corresponding to the first keyword from the records stored in the properties database; and

in the record corresponding to the first keyword, revising the public preference record data or the public dislike record data included in the corresponding record according to a user'"'"'s preference indication corresponding to the first keyword included in the first user'"'"'s preference data.

52. The language dialogue method as recited in claim 51, wherein if the one of the records corresponding to the first keyword is not searched from the properties database, a new record is established in the properties database according to the first keyword, and the public preference record data corresponding to the new record and the public dislike record data corresponding to the new record are revised according to the user'"'"'s preference indication corresponding to the first keyword.

53. The language dialogue method as recited in claim 37, further comprising:

parsing a second speech input to obtain at least one second keyword, wherein the second speech input is input by a user after the first speech response is output; and

selecting a second report answer other than the at least one first report answer from the candidate list according to the second keyword.

54. The language dialogue method as recited in claim 53, further comprising the language comprehension system captures a second user'"'"'s preference data from the second speech input and revises the user'"'"'s preference data stored into the properties database according to the second user'"'"'s preference data.

55. The language dialogue method as recited in claim 37, wherein the step of revising the user'"'"'s preference data stored into the properties database according to the first user'"'"'s preference data comprises:

searching one of the records corresponding to the first keyword from the records stored in the properties database; and

in the record corresponding to the first keyword, revising the stored user'"'"'s preference data corresponding to the first keyword according to a user'"'"'s preference indication corresponding to the first keyword included in the first user'"'"'s preference data.

56. The language dialogue method as recited in claim 55, further comprising if the one of the records corresponding to the first keyword is not searched from the properties database, a new record is established in the properties database according to the first keyword, and the user'"'"'s preference record data corresponding to the new record is revised according to the user'"'"'s preference indication corresponding to the first keyword.

57. A method of providing a corresponding report answer corresponding to user'"'"'s speech inputs in response to a user'"'"'s habit, the method being suitable for a mobile communication apparatus and comprising:

receiving a first speech input from a user, wherein the first speech input is a sentence comprising a noun of the sentence as a keyword and a preference indication; and

outputting a speech response that is a speech converted by the corresponding report answer in response to a user'"'"'s preference data stored by the user, the user'"'"'s preference corresponding to the keyword, wherein the speech response is output according to a most popular user'"'"'s preference of the user'"'"'s preference, wherein the first speech response is a sentence outputted through a speech output interface, wherein each of the report answers has a priority, the priorities of the report answers of the candidate list are determined according to usages of the report answers, the first report answer is one of the report answers with the highest priority,

wherein when a plurality of speech inputs other than the first speech input present the at least one report answer is incorrect, the corresponding report answer is removed from a candidate list which is obtained by parsing the keyword and executing a full-text search in a structured database according to the keyword,

wherein structured database comprises a plurality of records, each of the records comprises a title field, the title field comprises at least one sub-field, each of the at least one sub-field comprises an indication field and a value field, the indication field stores indication data, and the value field stores the value data,

wherein the full-text search compares the value data of each of records and the first keyword to determine whether each of the records is matched and selects the report answers by the matched records,

wherein the user'"'"'s preference data is revised according to the preference indication included in the first speech input.

58. The method as recited in claim 57, wherein the user'"'"'s preference data is one of the records indicating whether the user personally likes or dislikes the keyword.

59. The method as recited in claim 57, wherein the user'"'"'s preference is one of the records indicating whether the public likes or dislikes the keyword.

60. The method as recited in claim 57, the user'"'"'s preference data being stored into one of the records including a corresponding keyword, when the keyword included in the first speech input does not match the corresponding keyword of the record, establishing a new record and storing the preference indication included in the first speech input into the new record, wherein the keyword included in the first speech input acts as the corresponding keyword of the new record, and the preference indication stored in the new record acts as a new user'"'"'s preference.