Diagnosing voice application issues of an operational environment

• US 20060210025A1
• Filed: 03/18/2005
• Published: 09/21/2006
• Est. Priority Date: 03/18/2005
• Status: Active Grant

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Dhanakshirur, Girish
• Granted Patent: US 7,505,569 B2
• Time in Patent Office: Days
• US Class Current: 379/88.40
• CPC Class Codes: H04M 2201/60 Medium conversion
• CPC Class Codes: H04M 3/2227 Quality of service monitoring
• CPC Class Codes: H04M 3/493 Interactive information services, e.g. directory enquiries ; Arrangements therefor, e.g. interactive voice response [IVR] systems or voice portals

Abstract

A computerized method for diagnosing deployed voice-enabled application issues within a maintenance environment. The method can identify a recognition log file that includes details of previously occurring interactions of a voice-enabled application executing programmatic actions in an operational environment. The voice-enabled application can utilize a speech recognition service to convert utterances to text. Previously occurring interactions of the operational environment can be simulated in a maintenance environment. The maintenance environment can include a maintenance voice-enabled application component used in place of the voice-enabled application and can include a text-based speech recognition service used in place of the speech recognition service. The text-based speech recognition service can extract text from the recognition log file to simulate previously occurring utterance conversions performed by the speech recognition service without requiring the text-based speech recognition service to receive and convert utterances.

Claims

1. A computerized method for diagnosing deployed voice-enabled application issues within a maintenance environment comprising the steps of:

identifying a recognition log file that includes details of previously occurring interactions of a voice-enabled application executing programmatic actions in an operational environment, said voice-enabled application utilizing a speech recognition service to convert utterances to text;

simulating said previously occurring interactions in a maintenance environment, said maintenance environment including a maintenance voice-enabled application component used in place of the voice-enabled application and including a text-based speech recognition service used in place of the speech recognition service;

said text-based speech recognition service extracting text from the recognition log file to simulate previously occurring utterance conversions performed by the speech recognition service without requiring the text-based speech recognition service to receive and convert utterances.

2. The method of claim 1, where the previously occurring interactions are telephony interactions between a user and an interactive voice response system.

3. The method of claim 2, wherein the voice-enabled application comprises a voice browser that interfaces with the speech recognition service in a manner determined by voice-based markup code.

4. The method of claim 1, further comprising the steps of:

receiving an indication of an application problem occurring within the operational environment during a telephone interaction session for which the telephone interactions relate;

locating information within the recognition log file associated with the telephone interaction session; and

conveying the located information along with the indication to the maintenance environment to facilitate diagnosing the indicated application problem, wherein the extracted text comprises the located information.

5. The method of claim 1, wherein the speech recognition service performs utterance conversions for a plurality of applications, wherein said speech recognition service conveys tasks to a cluster of automated speech recognition engines, which are used to perform the utterance conversions.

6. The method of claim 1, wherein said recognition log file is a session-specific log file and wherein the previously occurring interactions are interactions of a communication session involving a user and the voice-enabled application.

7. The method of claim 6, further comprising the steps of:

in the operational environment, detecting a technical problem during the communication session;

responsive to the detecting, automatically conveying the session-specific log file along with a problem indication to the maintenance environment.

8. The method of claim 1, wherein the speech recognition service operates on a per-turn basis, wherein utterances are parsed into a plurality of turn-processable segments each segment processed by an automatic speech recognition engine.

9. The method of claim 8, wherein recognition log file records interactions on a per-turn basis, and wherein the text-based speech recognition service extracts text from the recognition log file on a per-turn basis.

10. The method of claim 1, wherein the recognition log file comprises for each logged utterance conversion event, at least one textual result and a corresponding confidence score for each textual result.

11. The method of claim 10, wherein the at least one textual result comprises a plurality of textual results, said plurality of textual results comprising a selected textual result and at least one alternative textual result.

12. The method of claim 11, wherein said plurality of textual results includes each result contained within an n-best list of results, wherein the n of the n-best results is a parameter for the number of text results returned per utterance conversion, said n-best results being those results having the greatest relative confidence scores.

13. The method of claim 11, wherein the plurality of textual results includes all textual results having an associated confidence score over a designated threshold value.

14. A text based recognition service component comprising:

a event receiving mechanism configured to receive recognition requests originating from a voice browser;

a recognition log reading mechanism configured to extract textual information from a recognition log file to simulate previously occurring utterance conversions; and

an event response mechanism configured provide a response for each received recognition request based upon the extracted textual information, wherein the text-based recognition service component is a component of a maintenance environment used in place of a recognition service component of an operational environment, wherein at least a portion of the extracted textual information was previously placed within the recognition log file by the recognition service component.

15. The text-based recognition service component of claim 14, wherein the recognition log file is a session log for an interaction session involving a user and the voice browser, wherein the voice browser interfaces with the speech recognition service in a manner determined by voice-based markup code, wherein the speech recognition service performs tasks on a per-turn basis, and wherein the recognition log file includes an entry for each turn.

16. A problem diagnostic method comprising the steps of:

in a session log, logging events handled by a speech recognition service for a deployed voice application during a telephony session, for each logged speech conversion event, said session log including at least one textual result and a corresponding confidence score for each textual result;

in a maintenance environment that simulates the telephony session, reaching a point in a telephony session simulation where the speech recognition service handled an event;

conveying an indication of the event to a replacement component for the speech recognition service;

when the conveyed indication is a speech conversion event, said replacement component extracting from the session log the at least one textual result and corresponding confidence score associated with the speech conversion event; and

utilizing the extracted data from the session log within the maintenance environment as a simulation equivalent of results that were generated by the speech recognition service.

17. The method of claim 16, wherein the at least one textual result includes a plurality of textual results, said plurality of textual results comprising a selected textual result and at least one alternative textual result.

18. The method of claim 17, wherein said plurality of textual results includes the n-best results generated for each utterance, wherein the n of the n-best results is a parameter for the number of text results returned per utterance conversion, said n-best results being those results having the greatest relative confidence scores.

19. The method of claim 17, wherein said plurality of textual results includes all results generated for an utterance that has a corresponding confidence score exceeding a designated threshold.

20. The method of claim 16, wherein the speech recognition service performs tasks on a per-turn basis, and wherein the session log includes an entry for each turn.