Method and system of generating reference variations for directory assistance data

• US 7,623,648 B1
• Filed: 12/01/2004
• Issued: 11/24/2009
• Est. Priority Date: 12/01/2004
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation), Silverbrook Research Pty Limited
• Original Assignee: Tellme Networks Incorporated ([24]7.ai, Inc.)
• Inventors: Kibre, Nick, Mitby, David, Oppenheim, Kyle
• Primary Examiner(s): Anwah; Olisa
• Application Number:
  US11/002,829
• Time in Patent Office: 1,819 Days
• Field of Search: 704/270, 704/257, 704/251, 704/252, 704/255, 704/231, 704/10, 379/218.01, 379/88.01
• US Class Current: 379/218.01
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: H04M 2201/40 using speech recognition speech recognition per se G10L15/00
• CPC Class Codes: H04M 3/4931 Directory assistance systems

Abstract

Methods and systems of performing user input recognition. A digital directory comprising listings is accessed. Metadata information is associated with individual listings describing the individual listings. The metadata information is modified to generate transformed metadata information. Therefore, the transformed metadata information is generated as a function of context information relating to a typical user interaction with the listings. Information is generated for aiding in an automated user input recognition process based on the transformed metadata information.

Claims

1. A method of performing user input recognition comprising:

accessing a digital directory comprising a plurality of listings;

associating individual listings with metadata information describing said individual listings;

modifying the metadata information to generate transformed metadata information, wherein the transformed metadata information is generated as a function of context information relating to a typical user interaction with the listings;

wherein the modifying the metadata information comprises identifying groups of listings as pertaining to recognized listing group types;

generating information for aiding in an automated user input recognition process based on said transformed metadata information.

2. A method as described in claim 1 wherein said automated user input recognition process is a speech recognition process.

3. A method as described in claim 2 wherein said information for aiding in an automated user input recognition process comprises a grammar.

4. A method as described in claim 1 wherein said automated user input recognition process is a text input recognition process.

5. A method as described in claim 1 wherein said generating information for aiding in an automated user input recognition process comprises:

based on said transformed metadata information, generating a plurality of reference variances associated with the listings; and

based on said plurality of reference variances, generating a grammar as said information for aiding in an automated user input recognition process.

6. A method as described in claim 5 wherein said plurality of reference variances reflect typical variations in human references to the listings.

7. A method as described in claim 5 further comprising supplying said grammar to said automated user input recognition process.

8. A method as described in claim 1 wherein said associating individual listings with metadata information describing said individual listings comprises tagging said individual listings.

9. A method as described in claim 1 wherein said metadata information describing said individual listings is selected from the group comprising:

a known entity;

an address;

a business name;

a main listing;

a department; and

a person'"'"'s name.

10. A method as described in claim 1 wherein said recognized listings group types comprise:

a retail group type;

a hospital group type;

an airline group type; and

a government group type.

11. A system for user input recognition comprising:

a first metadata generator operable to generate metadata information describing individual listings of a digital directory and to associate said metadata information with said individual listings;

wherein said first metadata generator comprises a plurality of tagger modules operable to determine meanings of individual listings of said digital directory and to generate respective tags associated therewith;

a second metadata generator operable to modify the metadata information as a function of context information relating to a typical user interaction with the listings; and

logic operable to generate information for aiding in a user input recognition process based on transformed metadata information resulting from the second metadata generator.

12. The system of claim 11, wherein said second metadata generator is further operable to tag listings in said digital directory based on metadata context.

13. The system of claim 11, wherein said second metadata generator is further operable to determine relevant tags for releasable directory listings in said digital directory based on metadata context.

14. The system of claim 13, further comprising a presentation name generator operable to generate presentation names for said releasable listings.

15. The system of claim 11, wherein said second metadata generator comprises a plurality of context handlers operable to determine meanings of individual listings of said digital directory based on identifying a group listing type within which an individual listing resides.

16. The system of claim 11, wherein said user input is speech and further comprising a grammar generator operable to generate speech recognition grammar based on said information for aiding in a user input recognition process.

17. The system of claim 11, wherein said logic is further operable to generate said information for aiding in a user recognition process based on said metadata information generated by said first metadata generator.

18. The system of claim 11, further comprising a directory search engine that is able to determine a telephone number from said digital directory based on a selected entry of said information for aiding in a user input recognition process.

19. The system of claim 11 wherein said user input is text based input.

20. A computer readable medium having stored thereon instructions that when executed on a processor implement a method of generating input parameters usable in a user input recognition process, said method comprising:

accessing computer readable telephone directory data;

performing contextual analysis of said telephone directory data to generate metadata therefrom;

wherein said performing contextual analysis comprises applying telephone operator frequency data;

generating a plurality of synonym phrases based on said metadata resulting from said contextual analysis wherein said plurality of synonym phrases represent typical user input variations in accessing data related to said listings; and

generating said input parameters usable in user input recognition based on said plurality of synonym phrases.

21. The computer readable medium of claim 20, wherein said user input is speech and wherein further said input parameters comprise a grammar and further comprising:

using a speech recognition engine supplied with said grammar to generate a result based on an input user voice signal; and

outputting said result to a user as a playback confirmation.

22. The computer readable medium of claim 20, wherein said metadata is associated with listings of said telephone directory data.

23. The computer readable medium of claim 22, wherein said performing contextual analysis comprises modifying a first metadata based on a first listing of said telephone directory data based on relationship between a second metadata associated with a second listing.

24. The computer readable medium of claim 20, wherein said method further comprises:

recognizing a first synonym phrase of said plurality of synonym phrases in a user input directory request;

determining a telephone number from said telephone directory data based on said first synonym phrase; and

supplying said telephone number to said user.

25. The computer readable medium of claim 20, wherein said performing contextual analysis comprises determining relevant tags for releasable telephone listings in said telephone directory data based on contextual analysis of said metadata.

26. The computer readable medium of claim 25, wherein said method further comprises generating presentation names for said releasable telephone listings.

27. The computer readable medium of claim 25, wherein said generating said plurality of synonym phrases further comprises basing at least one of said synonym phrases on said relevant tags for said releasable listings.

28. A system for performing speech recognition comprising a processor coupled to a memory and wherein said memory comprises instructions for implementing a method of performing speech recognition, said method comprising:

accessing a digital directory comprising a plurality of listings wherein listings of said digital directory are organized according to a hierarchy of captions;

associating individual listings with metadata information describing said individual listings;

modifying the metadata information to generate transformed metadata information, wherein the transformed metadata information is generated as a function of context information relating to a typical user interaction with the listings;

wherein the modifying the metadata information comprises identifying captions of listings as pertaining to recognized caption types;

generating information for aiding in speech recognition based on said transformed metadata information;

wherein said generating information for aiding in speech recognition comprises;

based on said transformed metadata information, generating a plurality of synonym phrases associated with the listings;

based on said plurality of synonym phrases, generating a grammar for aiding in a speech recognition engine.

29. A system as described in claim 28 wherein said plurality of synonym phrases reflect typical variations in human references to the listings.

30. A system as described in claim 28 wherein said method further comprise supplying said grammar to said speech recognition engine.

31. A system as described in claim 28 wherein said associating individual listings with metadata information describing said individual listings comprises tagging said individual listings.

32. A system as described in claim 28 wherein said metadata information describing said individual listings is selected from the group comprising:

a known entity;

an address;

a business name;

a main listing;

a department; and

a person'"'"'s name.

33. A system as described in claim 28 wherein said recognized caption types comprise:

a retail type;

a hospital type;

an airline type; and

a government type.

34. A system for user input recognition, the system, comprising:

a first metadata generator operable to generate metadata information describing individual listings of a digital directory and to associate said metadata information with said individual listings;

a second metadata generator operable to modify the metadata information as a function of context information relating to a typical user interaction with the listings;

wherein said second metadata generator comprises a plurality of context handlers operable to determine meanings of individual listings of said digital directory based on identifying a group listing type within which an individual listing resides; and

logic operable to generate information for aiding in a user input recognition process based on transformed metadata information resulting from the second metadata generator.

35. A computer readable medium having stored thereon instructions that when executed on a processor implement a method of generating input parameters usable in a user input recognition process, said method comprising:

accessing computer readable telephone directory data;

performing contextual analysis of said telephone directory data to generate metadata therefrom;

wherein said metadata is associated with listings of said telephone directory data;

generating a plurality of synonym phrases based on said metadata resulting from said contextual analysis wherein said plurality of synonym phrases represent typical user input variations in accessing data related to said listings;

wherein said performing contextual analysis comprises modifying a first metadata based on a first listing of said telephone directory data based on relationship between a second metadata associated with a second listing;

generating said input parameters usable in user input recognition based on said plurality of synonym phrases.

36. A computer readable medium having stored thereon instructions that when executed on a processor implement a method of generating input parameters usable in a user input recognition process, said method comprising:

accessing computer readable telephone directory data;

performing contextual analysis of said telephone directory data to generate metadata therefrom;

generating a plurality of synonym phrases based on said metadata resulting from said contextual analysis wherein said plurality of synonym phrases represent typical user input variations in accessing data related to said listings;

wherein said performing contextual analysis comprises determining relevant tags for releasable telephone listings in said telephone directory data based on contextual analysis of said metadata;

generating said input parameters usable in user input recognition based on said plurality of synonym phrases.

37. A method of performing input recognition on a listing in a data source, the method, comprising:

generating metadata for the listing in the data source;

associating the metadata with the listing;

modifying the metadata information to generate transformed metadata by analyzing the metadata using a heuristic rule;

performing pattern matching on the transformed metadata using a set of pattern rules;

in response to detecting a match, generating information for the listing to aid in recognition of a user input.

38. The method of claim 37, wherein the information includes reference variations of the listing.

39. The method of claim 37, further comprising, extracting a caption set for the listing to generate the metadata.

40. The method of claim 37, wherein, the metadata includes a type portion and a data portion.

41. The method of claim 37, further comprising,

receiving the user input from a user;

using the information, determining a referenced entity in the user input; and

returning associated information of the referenced entity to the user.

42. The method of claim 41, wherein, the user input includes spoken utterance.

43. The method of claim 41, wherein, the user input includes a text reference.

44. The method of claim 41, wherein, the information is used for speech recognition.

45. The method of claim 41, wherein, the information for the listing is generated using semantic analysis.

46. A method of performing input recognition on a listing in a data source, the method, comprising:

generating metadata tags for the listing in the data source;

associating the metadata tags with the listing;

generating transformed metadata by modifying the metadata tags;

performing pattern matching on the transformed metadata using a set of pattern rules;

in response to detecting a match, generating synonym phrases for the listing to facilitate recognition of user input.

47. The method of claim 46, further comprising, performing contextual tag adjustment on the metadata tags to generate the transformed metadata.

48. The method of claim 46, further comprising, generating a grammar using the synonym phrases.

49. The method of claim 48, wherein, the grammar is stored in a grammar database coupled to a speech recognition engine.

50. A method of performing input recognition on a listing in a data source, the method, comprising:

generating metadata for the listing in the data source;

associating the metadata with the listing;

modifying the metadata information to generate transformed metadata by analyzing the metadata using a heuristic rule;

using the transformed metadata, generating reference variations for the listing to facilitate recognition of user input.