Systems and methods for providing automated directory assistance using transcripts

• US 7,401,023 B1
• Filed: 09/06/2000
• Issued: 07/15/2008
• Est. Priority Date: 09/06/2000
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: BBN Technologies (Rtx Corporation), Verizon Corporate Services Group Incorporated (Verizon Communications Inc.)
• Inventors: Makhoul, John, Shu, Han, Schwartz, Richard Mark, Nguyen, Long
• Primary Examiner(s): Opsasnick, Michael N
• Application Number:
  US09/656,264
• Time in Patent Office: 2,869 Days
• Field of Search: 704/270, 704/275, 704/277, 379/218.01, 379/218.02
• US Class Current: 704/275
• 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

An automated directory assistance (130) includes a training system (210) and a directory assistance system (220). The training system (210) trains and maintains the directory assistance system (220). The training system (210) includes a transcription module (310), a speech grammar estimation module (330), a listings statistics estimation module (340), and a required words determination module (350). The transcription module (310) obtains transcripts relating to directory service requests. The speech grammar estimation module (330) creates an n-gram grammar for multiple telephone numbers from the transcripts. The listings statistics estimation module (340) identifies words used to refer to each of the telephone numbers from the transcripts. The required words determination module (350) identifies at least one word that is required to request each of the telephone numbers from the transcripts. The directory assistance system (210) includes a speech recognition module (510), a listing retrieval module (520), and an accept/reject module (530). The speech recognition module (510) receives an audible request for a telephone number from a caller and generates a transcript from the audible request. The listing retrieval module (520) retrieves at least one listing corresponding to the audible request from a database using the transcript. The accept/reject module (530) determines whether to accept one or more of the listings retrieved by the listing retrieval module (520) and presents a telephone number corresponding to the accepted listing to the caller.

Claims

1. An automated directory assistance system, comprising:

a speech recognition module configured to receive an audible request for a telephone number from a caller and generate a transcript from the audible request;

a listing retrieval module configured to retrieve a listing corresponding to the audible request from a database, the listing retrieval module being configured to use the transcript as a query, the database storing words previously included in requests for telephone numbers; and

an accept/reject module configured to determine whether to accept the listing retrieved by the listing retrieval module and present a telephone number corresponding to the accepted listing to the caller, the listing being rejected unless the transcript contains at least one of the stored words.

2. The system of claim 1, wherein the speech recognition module includes a large vocabulary speech recognizer.

3. The system of claim 1, wherein the speech recognition module is configured to use acoustic models and an n-gram grammar to recognize at least one word included in the audible request.

4. The system of claim 1, wherein the listing retrieval module includes a statistical information retrieval system.

5. The system of claim 1, wherein the listing retrieval module is configured to rank a plurality of listings from the database.

6. The system of claim 1, wherein the accept/reject module is configured to identify at least one word that is required for the listing, determine whether the transcript contains the identified at least one required word, and accept the listing when the transcript contains the identified at least one required word.

7. The system of claim 1, wherein the accept/reject module is further configured to transfer the audible request to a human operator when the accept/reject module does not accept the listing.

8. The system of claim 1, further comprising:

a training system configured to automatically configure the speech recognition module, the listing retrieval module, and the accept/reject mode.

9. The system of claim 8, wherein the training system includes:

an acoustic model training module configured to estimate acoustic models from training transcripts relating to previous requests for the telephone numbers, the telephone numbers including the telephone number corresponding to the listing; and

a speech grammar estimation module configured to create an n-gram grammar for the telephone numbers, the speech recognition module using the acoustic models and the n-gram grammar to generate the transcript from the audible request.

10. The system of claim 8, wherein the training system includes:

a listings statistics estimation module configured to identify words used when requesting the telephone number, the listing retrieval module retrieving the identified words using the transcript.

11. The system of claim 8, wherein the training system includes:

a required words determination module configured to identify at least one word that is required to request the telephone number, the accept/reject module using the identified at least one required word to determine whether the listing is acceptable.

12. The system of claim 8, wherein the training system includes:

a transcription module configured to automatically generate training transcripts corresponding to previous requests for the telephone numbers, the telephone numbers including the telephone number.

13. The system of claim 12, wherein the transcription module includes:

a grammar creation component configured to create a loose grammar corresponding to each of the telephone numbers, and

a speech recognition component configured to generate a training transcript for one of the previous requests for the telephone numbers using the loose grammar.

14. The system of 13, wherein the transcription module further includes:

an accept/reject component configured to determine whether the generated training transcript is acceptable.

15. The system of claim 14, wherein the transcription module further includes:

a verification/correction module configured to present the generated training transcript to a human for at least one of verification and modification.

16. A method for providing directory assistance, comprising:

receiving an audible request for a telephone number from a caller;

generating a transcript from the audible request;

retrieving a listing corresponding to the audible request from a database using the transcript as a query into the database, the database storing words previously included in requests for telephone numbers;

determining whether to accept the retrieved listing, the listing being rejected unless the transcript contains at least one of the stored words; and

presenting a telephone number corresponding to the accepted listing to the caller.

17. The method of claim 16, comprising:

estimating acoustic models from speech and training relating to requests for the telephone numbers; and

creating an n-gram grammar for the telephone numbers.

18. The method of claim 17, wherein the generating includes:

using the acoustic models and the grammar to recognize at least one word included in the audible request.

19. The method of claim 16, further comprising:

identifying words relating to each of the telephone numbers, the telephone numbers including the telephone number; and

storing the words in the database.

21. The method of claim 16, wherein the retrieving includes:

retrieving a plurality of listings from the database, the listings corresponding to the audible request; and

ranking the plurality of listings.

22. The method of claim 16, wherein the determining includes:

identifying at least one word that is required for the listing,

determining whether the transcript contains the identified at least one required word, and

accepting the retrieved listing when the transcript contains the identified at least one required word.

23. The method of claim 16, comprising:

retrieving a plurality of listings corresponding to the audible request; and

transferring the audible request to a human operator when none of the plurality of retrieved listings are accepted.

24. The method of claim 16, further comprising:

automatically generating training transcripts corresponding to previous requests for the telephone numbers, the telephone numbers including the telephone number.

20. The method of 19, wherein the retrieving includes:

using the transcript as a query into database to retrieve the words relating to the telephone number.

25. The method 24, wherein the automatically generating includes:

creating a loose grammar corresponding to each of the telephone numbers, and

generating a training transcript for one of the previous requests for one of the telephone numbers using the loose grammar.

26. The method of claim 25, wherein the automatically generating includes:

determining whether the generated transcript is acceptable.

27. The method of claim 26, wherein the automatically generating further includes:

presenting the transcript to a human for at least one of verification and modification.

28. A system for providing automated directory assistance, comprising:

means for receiving a request for a telephone number from a caller;

means for generating a transcript from the request;

means for retrieving a listing corresponding to the request from a database using the transcript as a query, the database storing words previously included in requests for telephone numbers, the query including matching the transcript to the stored words;

means for determining whether to accept the retrieved listing; and

means for presenting a telephone number corresponding to the accepted listing to the caller.

29. A computer-readable medium that stores instructions executable by at least one processor to perform a method for providing directory assistance, comprising:

recognizing at least one word in an audible request for a telephone number received from a caller;

generating a transcript from the audible request;

retrieving at least one listing corresponding to the audible request from a database using the transcript as a query, the database storing words previously included in requests for telephone numbers, the query including matching the transcript to the stored words;

determining whether to accept one or more of the retrieved listings; and

presenting a telephone number corresponding to the accepted one or more listings to the caller.

30. A method for providing a directory assistance service, comprising:

receiving a request for a telephone number from a caller, the request being spoken by the caller and including a location and listing corresponding to the telephone number;

using large vocabulary speech recognition to recognize at least one word spoken by the caller when making the request;

generating a transcript from the at least one word;

using statistical information retrieval and the transcript to identify a listing corresponding to the recognized word, including using the transcript as a query, the listings database storing words previously included in requests for telephone numbers, the query including matching the transcript to the stored words;

determining whether the listing is likely to be correct; and

providing a telephone number corresponding to the listing to the caller.

31. A method for providing a directory assistance service, comprising:

receiving a request for a telephone number from a caller, the request being spoken by caller and including a location and listing corresponding to the telephone number;

using large vocabulary speech recognition to recognize at least one word spoken by the caller when making the request;

generating a transcript from at least one word;

using statistical information retrieval and the transcript to identity a listing corresponding to the recognized word, the statistical information retrieval matching the transcript to words previously included in requests for telephone numbers; and

connecting the caller to a called party corresponding to the listing.

32. A method of providing directory assistance, comprising:

defining a set of words or phrases associated with a listing;

defining at least one required word associated with the listing;

receiving a request for a telephone number from a caller;

using speech recognition to generate a transcript from the audible request;

using the transcript to identify the listing as a potential match to the request, the listing being identified by using the transcript as a query into a database containing the set of words or phrases associated with the listing, the set of words or phrases associated with the listing being previously included in requests for telephone numbers;

determining whether the transcript includes the at least one required word associated with the listing;

accepting the listing when it is that the transcript includes a match for at least one of the at least one required word; and

rejecting the listing when it is determined that the transcript does not include a match for at one of the at least one required word.