Granted Patent B2

US 8,645,136 · App. 12/840,190 · Granted Feb 4, 2014

Abstract

Claims (61)

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

a parser configured to parse voice streams from calls into utterances and assign an initial transcribed value and corresponding recognition score to each utterance within the voice streams;

a message generator configured to generate, for each call, a transcribed message comprising some of the initial transcribed values;

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

a comparison module configured to compare at least one questionable utterance from one call to other questionable utterances from other calls, to identify, within a predetermined time to successfully identify an appropriately sized pool of similar questionable utterances, a predetermined number of the other questionable utterances that are assigned initial transcribed values similar to at least one initial transcribed value which is assigned to the at least one questionable utterance, and to combine the identified other questionable utterances with the at least one questionable utterance into a group;

an assignment module configured to select a sample of the questionable utterances in the group, to receive a common manual transcription value for the selected sample of questionable utterances, and to assign the common manual transcription value to the remaining questionable utterances in the group; and

a processor configured to execute the parser, message generator, and modules.

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

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

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

a pool generator configured to build the group 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 configured to select the sample 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 configured to determine a similarity between the at least one questionable utterance and the other questionable utterances based on at least one of an initial transcribed value, recognition score, range of similarity, and characteristics of a call.

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

parsing voice streams from calls into utterances and assigning an initial transcribed value and corresponding recognition score to each utterance within the voice streams;

generating, for each call, a transcribed message for each call comprising some of 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 from one call to other questionable utterances from other calls;

identifying, within a predetermined time to successfully identify an appropriately sized pool of similar questionable utterances, a predetermined number of other questionable utterances that are assigned initial transcribed values similar to at least one initial transcribed value which is assigned to the at least one questionable utterance and combining the identified other questionable utterances with the at least one questionable utterance into a group;

selecting a sample of the questionable utterances from the group and receiving a common manual transcription value for the selected sample of questionable utterances; and

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

wherein the steps are performed by a processor.

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

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

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

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

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

selecting the sample 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 an initial transcribed value, recognition score, range of similarity, and characteristics of a call.

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

a parser configured to parse calls into speech utterances and assign an initial transcribed value and corresponding recognition score to each utterance;

a message generator configured to generate, for each call, a transcribed message comprising some of the initial transcribed values and some of the recognition scores;

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

a sample module configured to generate a sample for at least one questionable utterance from one call by identifying, within a predetermined time to successfully identify an appropriately sized pool of similar questionable utterances, other questionable utterances from other calls that are assigned initial transcribed values similar to at least one initial transcribed value which is assigned to the at least one questionable utterance, by combining the at least one questionable utterance with the identified other questionable utterances into a group, and by selecting a portion of the group as the sample using at least one of random sampling, specific selection sampling, and a combination of random and specific selection sampling;

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

a processor configured to execute the parser, message generator, and modules.

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

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

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

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

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

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

15. A method for hybrid voice transcription error reduction, comprising the steps of:

parsing calls into speech utterances and assigning an initial transcribed value and corresponding recognition score to each utterance;

generating, for each call, a transcribed message comprising some of the initial transcribed values and some of the corresponding recognition scores;

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

generating a sample for at least one questionable utterance from one call, comprising:

identifying within a predetermined time to successfully identify an appropriately sized pool of similar questionable utterances, the other questionable utterances from other calls that are assigned initial transcribed values similar to at least one initial transcribed value which is assigned to the at least one questionable utterance;

combining the at least one questionable utterance with the identified other questionable utterances into a group; and

selecting a portion of the group as the sample using at least one of random sampling, specific selection sampling, and a combination of random and specific selection sampling; and

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

wherein the steps are performed by a suitably-programmed computer.

16. A method according to claim 15 , further comprising:

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

17. A method according to claim 15 , further comprising:

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

18. A method according to claim 15 , further comprising:

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

Assignments (22)