STORED PHRASE REUTILIZATION WHEN TESTING SPEECH RECOGNITION

• US 20090216533A1
• Filed: 02/25/2008
• Published: 08/27/2009
• Est. Priority Date: 02/25/2008
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: PALGON, JONATHAN, LEITE, LEA T.
• Granted Patent: US 8,949,122 B2
• Time in Patent Office: Days
• US Class Current: 704/251
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/193 Formal grammars, e.g. finite state automata, context free grammars or word networks
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

A set of audio phrases and corresponding phrase characteristics can be maintained, such as in a database. The phrase characteristics can include a translation of speech in the associated audio phrase. A finite state grammar that includes a set of textual phrases can be received. A software algorithm can execute to compare the set of textual phrases against the translations associated with the maintained audio phrases. A result of the software algorithm execution can be produced, where the result indicates phrase coverage for the finite state grammar based upon the audio phrases.

Claims

1. A method for maximizing a use of stored phrases for speech recognition purposes comprising:

maintaining a plurality of audio phrases and corresponding phrase characteristics, wherein said phrase characteristics comprise a translation of speech in the associated audio phrase;

receiving at least one finite state grammar comprising a plurality of textual phrases;

executing a software algorithm to compare the plurality of textual phrases against the translations associated with the maintained audio phrases; and

producing a result of the software algorithm execution, where the result indicates phrase coverage for the finite state grammar based upon the audio phrases.

2. The method of claim 1, further comprising:

generating a test run package for the finite state grammar based upon the produced result; and

testing a speech recognition performance of the finite state grammar using the test run package.

3. The method of claim 1, further comprising:

integrating the software algorithm to a software development interface configured to create speech-enabled applications, wherein the finite state grammar is a speech recognition grammar associated with a voice user interface (VUI) being developed using the software development interface, and wherein the result is used generate a test run package comprising audio phrases and corresponding translations, which is used to test the speech recognition grammar.

4. The method of claim 1, wherein the at least one finite state grammar comprises a plurality of finite state grammars, said method further comprising:

maximizing a use of the plurality of audio phrases by utilizing each possible audio phrase which is valid for each of the plurality of finite state grammars when testing speech recognition performance of each of the plurality of finite state grammars.

5. The method of claim 1, wherein said phrase characteristics comprise at least two of a gender of a speaker, an age of a speaker, a native language of a speaker, a region where the speaker grew, at least one environmental characteristic of the audio phrase, at least one audio capture characteristic of the audio phrase, and a signal to noise ratio (SNR) present in the audio phrase.

6. The method of claim 5, further comprising:

specifying at least one constraint for the software algorithm, wherein said constraint specifies a value for at least one of the phrase characteristics;

limiting the comparison of the software algorithm to only those audio phrases having an associated phrase characteristic value matching the value of the specified constraint.

7. The method of claim 5, wherein said result is a user configured report, wherein said report comprises summary statistics relating to the phrase coverage.

8. The method of claim 1, further comprising:

detecting a programmatically triggered event, which causes the software algorithm to automatically execute and to produce the result, wherein the programmatically triggered event results from a detection of at least one of a change in the finite state grammar and a change in the plurality of audio phrases.

9. A computer program product for maximizing a use of stored phrases for speech recognition purposes, the computer program product comprising:

a computer usable medium having computer usable program code embodied therewith, the computer usable program code comprising;

computer usable program code configured to maintain a plurality of audio phrases and corresponding phrase characteristics, wherein said phrase characteristics comprise a translation of speech in the associated audio phrase;

computer usable program code configured to receive at least one finite state grammar comprising a plurality of textual phrases;

computer usable program code configured to execute a software algorithm to compare the plurality of textual phrases against the translations associated with the maintained audio phrases; and

computer usable program code configured to produce a result of the software algorithm execution, where the result indicates phrase coverage for the finite state grammar based upon the audio phrases.

10. The computer program product of claim 9, further comprising:

computer usable program code configured to generate a test run package for the finite state grammar based upon the produced result; and

computer usable program code configured to test a speech recognition performance of the finite state grammar using the test run package.

11. The computer program product of claim 9, further comprising:

computer usable program code configured to maximize a use of the plurality of audio phrases by utilizing each possible audio phrase which is valid for each of the plurality of finite state grammars when testing speech recognition performance of each of the plurality of finite state grammars.

12. The computer program product of claim 9, wherein said phrase characteristics comprise at least two of a gender of a speaker, an age of a speaker, a native language of a speaker, a region where the speaker grew, at least one environmental characteristic of the audio phrase, at least one audio capture characteristic of the audio phrase, and a signal to noise ratio (SNR) present in the audio phrase.

13. The computer program product of claim 12, further comprising:

computer usable program code configured to specify at least one constraint for the software algorithm, wherein said constraint specifies a value for at least one of the phrase characteristics;

computer usable program code configured to limit the comparison of the software algorithm to only those audio phrases having an associated phrase characteristic value matching the value of the specified constraint.

14. A system for maximizing a set of stored phrases comprising:

at least one data store comprising digitally encoded information specifying a plurality of audio phrases and associated phrase characteristics, wherein said phrase characteristics comprises a translation of speech in the associated audio phrase;

a coverage software tool configured to receive as input an identifier for a finite state grammar, wherein said finite state grammar comprises a plurality of textual phrases associated with a voice user interface (VUI), said coverage software tool being further configured to compare the plurality of textual phrases against the translations associated with the maintained audio phrases and to producing a result, wherein said result indicates phrase coverage for the finite state grammar based upon the audio phrases.

15. The system of claim 14, further comprising:

a recognition grammar tester configured to test a speech recognition performance of the finite state grammar based upon the plurality of audio phrases of the data store which the result indicates correspond to at least one of the textual phrases.

16. The system of claim 14, wherein said coverage tool is configured to maximize a use of the plurality of audio phrases by utilizing each possible audio phrase which is valid for each of the plurality of finite state grammars when testing speech recognition performance of each of the plurality of finite state grammars.

17. The system of claim 14, wherein said phrase characteristics comprise at least two of a gender of a speaker, an age of a speaker, a native language of a speaker, a region where the speaker grew, at least one environmental characteristic of the audio phrase, at least one audio capture characteristic of the audio phrase, and a signal to noise ratio (SNR) present in the audio phrase, and wherein the coverage software tool permits constraints to be specified which limit a comparison to a defined set of phrases having a value for at least one of the phrase characteristics matching a value defined in a constraint.

18. The system of claim 14, wherein the coverage tool further comprises:

an invocation engine configured to trigger an execution of a comparison action between a finite state grammar and said audio phrases;

a comparator configured to perform the comparison; and

a report generator configured to define presentation characteristics of the result.

19. The system of claim 18, wherein said invocation engine is configured to automatically trigger the execution of the comparison responsive to at least one of a change in the audio phrases of the data store and a change to the textual phrases in the finite state grammar.

20. The system of claim 18, wherein the report generator is configured to structure query language (SQL) defined reports, based upon user configured settings, and wherein the report generator is configured to include within the result summary statistics relating to the phrase coverage.