Method and apparatus for speech processing incorporating user intent

• US 7,647,228 B2
• Filed: 02/14/2006
• Issued: 01/12/2010
• Est. Priority Date: 11/03/2005
• Status: Active Grant

• Current Assignee: HTC Corporation
• Original Assignee: Apptera Incorporated
• Inventors: Silvera, M. Marketta, Chiu, Leo
• Primary Examiner(s): MCFADDEN, SUSAN IRIS
• Application Number:
  US11/276,114
• Publication Number:
  US 20070100625A1
• Time in Patent Office: 1,428 Days
• Field of Search: 704/257, 704/275
• US Class Current: 704/275
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 2015/223 Execution procedure of a spoken command

Abstract

An audible utterance is received from a user. A procedure determines a user intent associated with the audible utterance and identifies multiple parameters associated with the user intent. The procedure then determines whether any of the multiple parameters associated with the user intent are missing from the audible utterance. If any parameters are missing, an audible response requesting the missing parameters is generated and communicated to the user. If all parameters associated with the user intent are contained in the audible utterance, a response to the user'"'"'s request is generated and communicated to the user.

Claims

1. A method comprising:

receiving an audible utterance at a dialog processor;

determining a user intent associated with the audible utterance using an ontology, wherein the ontology includes a plurality of words that may be contained in audible utterances, and wherein the ontology defines relationships between the plurality of words that may be contained in audible utterances;

identifying a plurality of parameters associated with the user intent;

determining whether any of the plurality of parameters associated with the user intent are missing from the audible utterance;

if any of the plurality of parameters associated with the user intent are missing from the audible utterance, generating an audible response to the user requesting the missing parameters; and

if all parameters associated with the user intent are contained in the audible utterance, generating an audible response to the user'"'"'s request.

2. A method as recited in claim 1 further comprising converting the content of the audible utterance to a string of text characters after receiving the audible utterance from the user.

3. A method as recited in claim 1 wherein determining a user intent associated with the audible utterance includes identifying key words contained in the audible utterance.

4. A method as recited in claim 3 wherein determining a user intent associated with the audible utterance includes identifying a plurality of data elements contained in the ontology.

5. A method as recited in claim 1 further comprising:

identifying key words contained in the audible utterance;

identifying a plurality of data elements contained in the ontology;

comparing the identified key words with the plurality of data elements contained in the ontology; and

if the identified key words match a set of data elements contained in the ontology, associating that set of data elements with the user intent.

6. A method as recited in claim 1 further comprising generating an audible response to the user indicating that the user'"'"'s request was not understood if a user intent associated with the audible utterance could not be determined.

7. A method as recited in claim 1 wherein the audible utterance is received in the natural language of the user.

8. A method as recited in claim 1 wherein generating an audible response to the user'"'"'s request includes accessing at least one data source to obtain information necessary to respond to the user'"'"'s request.

9. A method as recited in claim 1 wherein determining a user intent associated with the audible utterance includes identifying a root word associated with words contained in the audible utterance.

10. A method comprising:

receiving an audible utterance at a dialog processor;

attempting to determine a user intent associated with the audible utterance using an ontology that includes a plurality of words that may be contained in audible utterances, wherein the ontology defines relationships between the plurality of words that may be contained in audible utterances;

if the user intent associated with the audible utterance is not determined, generating an audible response to the user indicating that the user'"'"'s request was not understood;

if the user intent associated with the audible utterance is determined;

identifying a plurality of parameters associated with the user intent;

determining whether any of the plurality of parameters associated with the user intent are missing from the audible utterance;

if any of the plurality of parameters associated with the user intent are missing from the audible utterance, generating an audible response to the user requesting the missing parameters; and

if all parameters associated with the user intent are contained in the audible utterance, generating an audible response to the user'"'"'s request.

11. A method as recited in claim 10 wherein attempting to determine a user intent associated with the audible utterance includes:

identifying key words contained in the audible utterance;

identifying any root words associated with the key words contained in the audible utterance; and

comparing the identified key words and root words with the plurality of words contained in the ontology.

12. A method as recited in claim 11 further comprising identifying the user intent by matching the identified key words and root words with a particular set of ontology data elements.

13. A method as recited in claim 10 wherein the audible utterance is spoken in the user'"'"'s natural language.

14. A method as recited in claim 10 wherein the audible utterance is received from a telephone operated by the user.

15. An apparatus comprising:

an ontology containing a plurality of data elements defining types of information that may be contained in user utterances, wherein the ontology defines relationships between words that may be contained in user utterances;

a speech grammar generator coupled to the ontology, the speech grammar generator to generate a natural language grammar;

a dialog manager coupled to the speech grammar generator, the dialog manager to receive user utterances and communicate responses to users;

an intent identification engine coupled to the dialog manager and the ontology, the intent identification engine to determine a user'"'"'s intent, wherein the user'"'"'s intent has a plurality of associated parameters;

a parameter qualifier coupled to the intent identification engine, the parameter qualifier to determine whether any parameters associated with a particular user'"'"'s intent are missing from the associated user utterance; and

a response generator coupled to the parameter qualifier, the response generator to generate responses to user utterances.

16. An apparatus as recited in claim 15 further comprising a data source coupled to the response generator, wherein the response generator accesses the data source when generating certain responses to user utterances.

17. An apparatus as recited in claim 15 wherein the dialog manager receives user utterances and communicates responses to users in the natural language of the user.

18. An apparatus as recited in claim 15 wherein the intent identification engine determines the user'"'"'s intent by matching a particular set of data elements in the ontology with key words contained in the user'"'"'s utterance.

19. An apparatus as recited in claim 15 further comprising a semantic factoring engine coupled to the dialog manager and the intent identification engine, the semantic factoring engine to identify key words contained in each user utterance and to identify root words associated with words contained in the each user utterance.