Computer-implemented system and method for voice transcription error reduction
A computer-implemented system and method for voice transcription error reduction is provided. Speech utterances are obtained from a voice stream and each speech utterance is associated with a transcribed value and a confidence score. Those utterances with transcription values associated with lower confidence scores are identified as questionable utterances. One of the questionable utterances is selected from the voice stream. A predetermined number of questionable utterances from other voice streams and having transcribed values similar to the transcribed value of the selected questionable utterance are identified as a pool of related utterances. A further transcribed value is received for each of a plurality of the questionable utterances in the pool of related utterances. A transcribed message is generated for the voice stream using those transcribed values with higher confidence scores and the further transcribed value for the selected questionable utterance.
1. A computer-implemented system for voice transcription error reduction, comprising:
a parser configured to obtain speech utterances from a voice stream and to assign each speech utterance with a transcribed value and a confidence score;
an utterance identifier configured to identify as questionable utterances, those utterances with transcription values associated with lower confidence scores;
an utterance selection module configured to select one of the questionable utterances and to generate a pool of related utterances comprising the selected questionable utterance and a predetermined number of questionable utterances, wherein the predetermined number of questionable utterances are assigned transcribed values similar to the transcribed value of the selected questionable utterance and are from other voice streams;
a transcription receipt module configured to receive a common transcribed value for a portion of the questionable utterances in the pool of related utterances;
a message generator configured to generate, using those transcribed values with high confidence scores and using the common transcribed value, a transcribed message; and
a processor configured to execute the parser, utterance identifier, utterance selection module, transcription receipt module, and message generator.
2. A system according to claim 1 , further comprising:
an assignment module configured to assign the common transcribed value to any questionable utterance which is in the pool of related utterances, which is not part of the portion, and which has been assigned a transcribed value that does not match the common transcribed value; and
a confirmation module configured to confirm the transcribed value of any questionable utterance, which is in the pool of related utterances, which is not part of the portion, and which has been assigned a transcribed value that matches the common transcribed value.
3. A system according to claim 1 , wherein, the message generator generates for each of the other voice streams, transcribed messages using the common transcribed value.
4. A system according to claim 1 , further comprising:
a time module configured to assign a predetermined time period for generating the pool of related utterances.
5. A system according to claim 4 , further comprising:
an utterance assignment module configured to assign the selected questionable utterance from the voice stream and questionable utterances from the other voice streams to an individual for manual transcription when the pool of related utterances is not generated within the predetermined time period.
6. A system according to claim 4 , further comprising:
a time determination module configured to provide a short time period for the predetermined time period when the voice stream is processed during a live call and to provide for the predetermined time period, a time period that is longer than the short time period of the live call when the voice stream is processed as a recording of the live call.
7. A system according to claim 1 , wherein the common transcribed value is assigned by a human transcriber.
8. A system according to claim 1 , further comprising:
a range of transcribed values; and
an utterance assignment module configured to identify as the predetermined number of questionable utterances, questionable utterances that are from the other voice streams and that have been assigned transcribed values that fall within the range.
9. A system according to claim 1 , further comprising:
a number determination module configured to set the predetermined number as a lower number when the voice stream is processed during a live call and to set the predetermined number as a higher number when the voice stream is processed as a recording of the live call.
10. A computer-implemented method for voice transcription error reduction, comprising the steps of:
obtaining speech utterances from a voice stream and assigning each speech utterance with a transcribed value and a confidence score;
identifying as questionable utterances, those utterances with transcription values associated with lower confidence scores;
selecting one of the questionable utterances and generating a pool of related utterances comprising the selected questionable utterance and a predetermined number of questionable utterances, wherein the predetermined number of questionable utterances are assigned transcribed values similar to the transcribed value of the selected questionable utterance and are from other voice streams;
receiving a common transcribed value for a portion of the questionable utterances in the pool of related utterances; and
generating using those transcribed values with high confidence scores and using the common transcribed value, a transcribed message,
wherein the steps are performed by a suitably-programmed computer.
11. A method according to claim 10 , further comprising:
assigning the common transcribed value to any questionable utterance which is in the pool of related utterances, which is not part of the portion, and which has been assigned a transcribed value that does not match the common transcribed value; and
confirming the transcribed value of any questionable utterance, which is in the pool of related utterances, which is not part of the portion, and which has been assigned a transcribed value that matches the common transcribed value.
12. A method according to claim 10 , further comprising:
generating for each of the other voice streams, transcribed messages using the common transcribed value.
13. A method according to claim 10 , further comprising:
assigning a predetermined time period for generating the pool of related utterances.
14. A method according to claim 13 , further comprising:
assigning the selected questionable utterance from the voice stream and questionable utterances from the other voice streams to an individual for manual transcription when the pool of related utterances is not generated within the predetermined time period.
15. A method according to claim 13 , further comprising at least one of:
providing a short time period for the predetermined time period when the voice stream is processed during a live call; and
providing for the predetermined time period, a time period that is longer than the short time period of the live call when the voice stream is processed as a recording of the live call.
16. A method according to claim 10 , wherein the common transcribed value is assigned by a human transcriber.
17. A method according to claim 10 , further comprising:
determining a range of transcribed values; and
identifying as the predetermined number of questionable utterances, questionable utterances that are from the other voice streams and that have been assigned transcribed values that fall within the range.
18. A method according to claim 10 , further comprising at least one of:
setting the predetermined number as a lower number when the voice stream is processed during a live call; and
setting the predetermined number as a higher number when the voice stream is processed as a recording of the live call.