System and Method for Efficiently Reducing Transcription Error Using Hybrid Voice Transcription

• US 20120022865A1
• Filed: 07/20/2010
• Published: 01/26/2012
• Est. Priority Date: 07/20/2010
• Status: Active Grant

• Current Assignee: Avaya LLC (Avaya Incorporated), Avaya Management L.P. (Avaya Incorporated)
• Original Assignee: Intellisist, Inc. dba Spoken Communications (Avaya Incorporated)
• Inventors: Milstein, David
• Granted Patent: US 8,645,136 B2
• Time in Patent Office: Days
• US Class Current: 704/235
• CPC Class Codes: G10L 15/01 Assessment or evaluation of speech recognition systems
• CPC Class Codes: G10L 15/06 Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice G10L15/14 takes precedence
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/10 using distance or distortion measures between unknown speech and reference templates
• CPC Class Codes: G10L 15/1822 Parsing for meaning understanding
• CPC Class Codes: G10L 15/19 Grammatical context, e.g. disambiguation of the recognition hypotheses based on word sequence rules
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 2015/0631 Creating reference templates; Clustering
• CPC Class Codes: G10L 2015/0638 Interactive procedures
• CPC Class Codes: H04M 3/42221 Conversation recording systems at the subscriber's set H04M1/656
• CPC Class Codes: H04M 3/51 Centralised call answering arrangements requiring operator intervention , e.g. call or contact centers for telemarketing
• CPC Class Codes: H04M 3/5183 Call or contact centers with computer-telephony arrangements

Abstract

A system and method for efficiently reducing transcription error using hybrid voice transcription is provided. A voice stream is parsed from a call into utterances. An initial transcribed value and corresponding recognition score are assigned to each utterance. A transcribed message is generated for the call and includes the initial transcribed values. A threshold is applied to the recognition scores to identify those utterances with recognition scores below the threshold as questionable utterances. At least one questionable utterance is compared to other questionable utterances from other calls and a group of similar questionable utterances is formed. One or more of the similar questionable utterances is selected from the group. A common manual transcription value is received for the selected similar questionable utterances. The common manual transcription value is assigned to the remaining similar questionable utterances in the group.

Claims

1. A system for efficiently reducing transcription error via hybrid voice transcription, comprising:

a parser to parse a voice stream from a call into utterances and assigning an initial transcribed value and corresponding recognition score to each utterance;

a message generator to generate a transcribed message for the call comprising the initial transcribed values;

a threshold module to apply a threshold to the recognition scores to identify those utterances with recognition scores below the threshold as questionable utterances;

a comparison module to compare at least one questionable utterance to other questionable utterances from other calls and forming a group of similar questionable utterances; and

an assignment module to select one or more of the similar questionable utterances from the group, to receive a common manual transcription value for the selected similar questionable utterances, and to assign the common manual transcription value to the remaining similar questionable utterances in the group.

2. A system according to claim 1, further comprising:

a replacement module to replace the initial transcribed value for the at least one questionable utterance with the common manual transcription value in the transcribed message.

3. A system according to claim 1, further comprising:

a pool generator to build the group of similar questionable utterances based on at least one of a size threshold and a time threshold.

4. A system according to claim 1, further comprising:

a sample module to select a sample comprising the one or more similar questionable utterances based on one of random sampling, specific sampling, and a combination of random and specific sampling.

5. A system according to claim 1, further comprising:

a similarity module to determine a similarity between the at least one questionable utterance and the other questionable utterances based on at least one of the initial transcribed value, recognition score, range of similarity, and characteristics of the call.

6. A method for efficiently reducing transcription error via hybrid voice transcription, comprising:

parsing a voice stream from a call into utterances and assigning an initial transcribed value and corresponding recognition score to each utterance;

generating a transcribed message for the call comprising the initial transcribed values;

applying a threshold to the recognition scores to identify those utterances with recognition scores below the threshold as questionable utterances;

comparing at least one questionable utterance to other questionable utterances from other calls and forming a group of similar questionable utterances;

selecting one or more of the similar questionable utterances from the group and receiving a common manual transcription value for the selected similar questionable utterances; and

assigning the common manual transcription value to the remaining similar questionable utterances in the group.

7. A method according to claim 6, further comprising:

replacing the initial transcribed value for the at least one questionable utterance with the common manual transcription value in the transcribed message.

8. A method according to claim 6, further comprising:

building the group of similar questionable utterances based on at least one of a size threshold and a time threshold.

9. A method according to claim 6, further comprising:

selecting a sample comprising the one or more similar questionable utterances based on one of random sampling, specific sampling, and a combination of random and specific sampling.

10. A method according to claim 6, further comprising:

determining a similarity between the at least one questionable utterance and the other questionable utterances based on at least one of the initial transcribed value, recognition score, range of similarity, and characteristics of the call.

11. A system for hybrid voice transcription error reduction, comprising:

a parser to parse a call into speech utterances;

a message generator to generate a transcribed message for the call by assigning an initial transcribed value and corresponding recognition score to each utterance;

a threshold module to apply a confidence threshold to the recognition scores and to select those utterances with initial transcribed values that fall below the threshold as questionable utterances;

a sample module to generate a sample for at least one of the questionable utterances comprising identifying other questionable utterances from other calls that are similar to the at least one questionable utterance, grouping the at least one questionable utterance and similar questionable utterances as related utterances, and selecting one or more of the related utterances as the sample; and

a module to receive a common manual transcription value for each of the related utterances in the sample and assigning the common manual transcription value to the remaining related utterances in the group.

12. A system according to claim 11, further comprising:

a pool generator to form the group of the related utterances by identifying a threshold number of the related utterances within a predetermined time limit.

13. A system according to claim 11, further comprising:

a replacement module to replace the initial transcribed value with the common manual transcription value for the at least one questionable utterance in the transcribed message.

14. A system according to claim 11, wherein the sample is selected using at least one of random sampling, specific selection sampling, and a combination of random and specific selection sampling.

15. A system according to claim 11, further comprising:

a similarity module to determine the similarity of the at least one questionable utterance and the other questionable utterances based on one or more of the initial transcribed value, recognition score, range of similarity between the related utterances, and characteristics of the call.

16. A method for hybrid voice transcription error reduction, comprising:

parsing a call into speech utterances and generating a transcribed message for the call by assigning an initial transcribed value and corresponding recognition score to each utterance;

applying a confidence threshold to the recognition scores and selecting those utterances with initial transcribed values that fall below the threshold as questionable utterances;

generating a sample for at least one of the questionable utterances, comprising;

identifying other questionable utterances from other calls that are similar to the at least one questionable utterance;

grouping the at least one questionable utterance and similar questionable utterances as related utterances; and

selecting one or more of the related utterances as the sample;

receiving a common manual transcription value for each of the related utterances in the sample and assigning the common manual transcription value to the remaining related utterances in the group.

17. A method according to claim 16, further comprising:

forming the group of the related utterances by identifying a threshold number of the related utterances within a predetermined time limit.

18. A method according to claim 16, further comprising:

replacing the initial transcribed value with the common manual transcription value for the at least one questionable utterance in the transcribed message.

19. A method according to claim 16, wherein the sample is selected using at least one of random sampling, specific selection sampling, and a combination of random and specific selection sampling.

20. A method according to claim 16, further comprising:

determining the similarity of the at least one questionable utterance and the other questionable utterances based on one or more of the initial transcribed value, recognition score, range of similarity between the related utterances, and characteristics of the call.