Back to results
US 11,238,884 B2Granted
Systems and methods for recording quality driven communication management
AssigneeRed Box Recorders Limited
InventorsSimon Jolly, Tony Commander, Kyrylo Zotkin
Life of this patent
Estimating the term…
Abstract
A system, method and non-transitory computer readable medium for providing call quality driven communication management wherein an audio data stream of a communication session having one or more utterances is processed to generate a transcript of the communication session. The generated transcript is analyzed to determine whether a quality of the audio data stream, and one or more quality improvement measures when one or more audio artifacts are determined to be present in the audio data stream.
Claims17
A method comprising:
- receiving an audio data stream of at least a portion of a communication session, the audio data stream comprising one or more utterances;
- processing the audio data stream to generate a transcript of the communication session, the transcript comprising candidate text for each utterance in the audio data stream along with confidence measures;
- analyzing the candidate text and confidence measures to determine a quality of the audio data stream, wherein analyzing the transcript comprises:
- tagging, as recognized text, each candidate text having a confidence measure meeting certain criteria, and tagging, as unrecognized text, each candidate text having a confidence measure that fails to meet the certain criteria;
- and
- calculating an audio quality score based on the ratio of recognized text to unrecognized text;
- and
- initiating one or more quality improvement measures when the audio quality score indicates that one or more audio artifacts are present in the audio data stream;
- and
- establishing an audio quality baseline by:
- retrieving a previously captured audio data stream of a previous communications session, the previously captured audio data stream comprising one or more previous utterances;
- processing the previously captured audio data stream to generate another transcript of the previous communication session, the another transcript comprising additional candidate text for each previous utterance in the previously captured audio data stream along with additional confidence measures;
- calculating an average confidence measure for the additional candidate text as the audio quality baseline;
- storing the audio quality baseline in association with the previously captured audio data stream;
- and
- comparing the audio quality score to the audio quality baseline to determine whether the audio quality score indicates that one or more audio artifacts are present in the audio data stream.
13 more elements
A call monitoring system comprising:
- a non-transitory storage medium having a plurality of instructions stored thereon;
- and
- at least one processor configure to execute the instructions to:
- receive an audio data stream of at least a portion of a communication session, the audio data stream comprising one or more utterances;
- process the audio data stream to generate a transcript of the communication session, the transcript comprising candidate text for each utterance in the audio data stream along with confidence measures;
- analyze the candidate text and confidence measures to determine a quality of the audio data stream, wherein the processor is configured to execute the instructions to:
- tag, as recognized text, each candidate text having a confidence measure meeting certain criteria, and tag, as unrecognized text, each candidate text having a confidence measure that fails to meet the certain criteria;
- and
- calculate an audio quality score based on the ratio of recognized text to unrecognized text;
- and
- initiate one or more quality improvement measures when the audio quality score indicates that one or more audio artifacts are present in the audio data stream;
- and
- establish an audio quality baseline by:
- retrieving a previously captured audio data stream of a previous communications session, the previously captured audio data stream comprising one or more previous utterances;
- processing the previously captured audio data stream to generate another transcript of the previous communication session, the another transcript comprising additional candidate text for each previous utterance in the previously captured audio data stream along with additional confidence measures;
- calculating an average confidence measure for the additional candidate text as the audio quality baseline;
- storing the audio quality baseline in association with the previously captured audio data stream;
- and
- compare the audio quality score to the audio quality baseline to determine whether the audio quality score indicates that one or more audio artifacts are present in the audio data stream.
16 more elements
A non-transitory computer-readable medium comprising a plurality of instructions, the instructions being executable by a processor to:
- receive an audio data stream of at least a portion of a communication session, the audio data stream comprising one or more utterances;
- process the audio data stream to generate a transcript of the communication session, the transcript comprising candidate text for each utterance in the audio data stream along with confidence measures;
- analyze the candidate text and confidence measures to determine a quality of the audio data stream, wherein the processor is configured to execute the instructions to:
- tag, as recognized text, each candidate text having a confidence measure meeting certain criteria, and tag, as unrecognized text, each candidate text having a confidence measure that fails to meet the certain criteria;
- and
- calculate an audio quality score based on the ratio of recognized text to unrecognized text;
- and
- initiate one or more quality improvement measures when the audio quality score indicates that one or more audio artifacts are present in the audio data stream;
- and
- establish an audio quality baseline by:
- retrieving a previously captured audio data stream of a previous communications session, the previously captured audio data stream comprising one or more previous utterances;
- processing the previously captured audio data stream to generate another transcript of the previous communication session, the another transcript comprising additional candidate text for each previous utterance in the previously captured audio data stream along with additional confidence measures;
- calculating an average confidence measure for the additional candidate text as the audio quality baseline;
- storing the audio quality baseline in association with the previously captured audio data stream;
- and
- comparing the audio quality score to the audio quality baseline to determine whether the audio quality score indicates that one or more audio artifacts are present in the audio data stream.
13 more elements
A method comprising:
- receiving an audio data stream of at least a portion of a communication session, the audio data stream comprising one or more utterances;
- processing the audio data stream to generate a transcript of the communication session, the transcript comprising candidate text for each utterance in the audio data stream along with confidence measures;
- analyzing the candidate text and confidence measures to determine a quality of the audio data stream, wherein analyzing the transcript comprises:
- tagging, as recognized text, each candidate text having a confidence measure meeting certain criteria, and tagging, as unrecognized text, each candidate text having a confidence measure that fails to meet the certain criteria;
- and
- calculating an audio quality score based on the ratio of recognized text to unrecognized text;
- and
- initiating one or more quality improvement measures when the audio quality score indicates that one or more audio artifacts are present in the audio data stream;
- and
- establishing a baseline audio quality index by:
- retrieving a plurality of previously captured audio data streams and corresponding metadata, the previously captured audio data streams each comprising one or more previous utterances;
- processing each previously captured audio data stream to generate additional transcripts, the additional transcripts comprising additional candidate text along with additional confidence measures;
- calculating a plurality of baseline audio quality scores as the average confidence measure the additional candidate text in each additional transcript;
- and
- storing the plurality of baseline audio quality scores in association with the corresponding metadata;
- selecting a particular baseline audio quality score from the baseline audio quality index based on a metadata of the audio stream;
- and
- comparing the audio quality score to the particular baseline audio quality score to determine whether the audio quality score indicates that one or more audio artifacts are present in the audio data stream.
15 more elements
A method comprising:
- receiving an audio data stream of at least a portion of a communication session, the audio data stream comprising one or more utterances;
- processing the audio data stream to generate a transcript of the communication session, the transcript comprising candidate text for each utterance in the audio data stream along with confidence measures;
- analyzing the candidate text and confidence measures to determine a quality of the audio data stream, wherein analyzing the transcript comprises:
- tagging, as recognized text, each candidate text having a confidence measure meeting certain criteria, and tagging, as unrecognized text, each candidate text having a confidence measure that fails to meet the certain criteria;
- and
- calculating an audio quality score based on the ratio of recognized text to unrecognized text;
- and
- initiating one or more quality improvement measures when the audio quality score indicates that one or more audio artifacts are present in the audio data stream;
- and
- establishing an audio quality baseline by:
- synthesizing an ideal audio data stream free of any audio quality artifacts, the ideal audio data stream comprising one or more clear utterances;
- processing the ideal audio data stream to generate another transcript, the another transcript comprising additional candidate text for each clear utterance in the ideal audio data stream along with additional confidence measures;
- calculating an average confidence measure for the additional candidate text as the audio quality baseline;
- and
- comparing the audio quality score to the audio quality baseline to determine whether the audio quality score indicates that one or more audio artifacts are present in the audio data stream.
12 more elements
A call monitoring system comprising:
- a non-transitory storage medium having a plurality of instructions stored thereon;
- and
- at least one processor configure to execute the instructions to:
- receive an audio data stream of at least a portion of a communication session, the audio data stream comprising one or more utterances;
- process the audio data stream to generate a transcript of the communication session, the transcript comprising candidate text for each utterance in the audio data stream along with confidence measures;
- analyze the candidate text and confidence measures to determine a quality of the audio data stream, wherein the processor is configured to execute the instructions to:
- tag, as recognized text, each candidate text having a confidence measure meeting certain criteria, and tag, as unrecognized text, each candidate text having a confidence measure that fails to meet the certain criteria;
- and
- calculate an audio quality score based on the ratio of recognized text to unrecognized text;
- and
- initiate one or more quality improvement measures when the audio quality score indicates that one or more audio artifacts are present in the audio data stream;
- and
- establish a baseline audio quality index by:
- retrieving a plurality of previously captured audio data streams and corresponding metadata, the previously captured audio data streams each comprising one or more previous utterances;
- processing each previously captured audio data stream to generate additional transcripts, the additional transcripts comprising additional candidate text along with additional confidence measures;
- calculating a plurality of baseline audio quality scores as the average confidence measure the additional candidate text in each additional transcript;
- and
- storing the plurality of baseline audio quality scores in association with the corresponding metadata;
- select a particular baseline audio quality score from the baseline audio quality index based on a metadata of the audio stream;
- and
- compare the audio quality score to the particular baseline audio quality score to determine whether the audio quality score indicates that one or more audio artifacts are present in the audio data stream.
18 more elements
A call monitoring system comprising:
- a non-transitory storage medium having a plurality of instructions stored thereon;
- and
- at least one processor configure to execute the instructions to:
- receive an audio data stream of at least a portion of a communication session, the audio data stream comprising one or more utterances;
- process the audio data stream to generate a transcript of the communication session, the transcript comprising candidate text for each utterance in the audio data stream along with confidence measures;
- analyze the candidate text and confidence measures to determine a quality of the audio data stream, wherein the processor is configured to execute the instructions to:
- tag, as recognized text, each candidate text having a confidence measure meeting certain criteria, and tag, as unrecognized text, each candidate text having a confidence measure that fails to meet the certain criteria;
- and
- calculate an audio quality score based on the ratio of recognized text to unrecognized text;
- and
- initiate one or more quality improvement measures when the audio quality score indicates that one or more audio artifacts are present in the audio data stream;
- and
- establish an audio quality baseline by:
- synthesizing an ideal audio data stream free of any audio quality artifacts, the ideal audio data stream comprising one or more clear utterances;
- processing the ideal audio data stream to generate another transcript, the another transcript comprising additional candidate text for each clear utterance in the ideal audio data stream along with additional confidence measures;
- and
- calculating an average confidence measure for the additional candidate text as the audio quality baseline;
- and
- compare the audio quality score to the audio quality baseline to determine whether the audio quality score indicates that one or more audio artifacts are present in the audio data stream.
16 more elements
A non-transitory computer-readable medium comprising a plurality of instructions, the instructions being executable by a processor to:
- receive an audio data stream of at least a portion of a communication session, the audio data stream comprising one or more utterances;
- process the audio data stream to generate a transcript of the communication session, the transcript comprising candidate text for each utterance in the audio data stream along with confidence measures;
- analyze the candidate text and confidence measures to determine a quality of the audio data stream, wherein the processor is configured to execute the instructions to:
- tag, as recognized text, each candidate text having a confidence measure meeting certain criteria, and tag, as unrecognized text, each candidate text having a confidence measure that fails to meet the certain criteria;
- and
- calculate an audio quality score based on the ratio of recognized text to unrecognized text;
- and
- initiate one or more quality improvement measures when the audio quality score indicates that one or more audio artifacts are present in the audio data stream;
- and
- establish a baseline audio quality index by:
- retrieving a plurality of previously captured audio data streams and corresponding metadata, the previously captured audio data streams each comprising one or more previous utterances;
- processing each previously captured audio data stream to generate additional transcripts, the additional transcripts comprising additional candidate text along with additional confidence measures;
- calculating a plurality of baseline audio quality scores as the average confidence measure the additional candidate text in each additional transcript;
- and
- storing the plurality of baseline audio quality scores in association with the corresponding metadata;
- select a particular baseline audio quality score from the baseline audio quality index based on a metadata of the audio stream;
- and
- compare the audio quality score to the particular baseline audio quality score to determine whether the audio quality score indicates that one or more audio artifacts are present in the audio data stream.
15 more elements
A non-transitory computer-readable medium comprising a plurality of instructions, the instructions being executable by a processor to:
- receive an audio data stream of at least a portion of a communication session, the audio data stream comprising one or more utterances;
- process the audio data stream to generate a transcript of the communication session, the transcript comprising candidate text for each utterance in the audio data stream along with confidence measures;
- analyze the candidate text and confidence measures to determine a quality of the audio data stream, wherein the processor is configured to execute the instructions to:
- tag, as recognized text, each candidate text having a confidence measure meeting certain criteria, and tag, as unrecognized text, each candidate text having a confidence measure that fails to meet the certain criteria;
- and
- calculate an audio quality score based on the ratio of recognized text to unrecognized text;
- and
- initiate one or more quality improvement measures when the audio quality score indicates that one or more audio artifacts are present in the audio data stream;
- and
- establish an audio quality baseline by:
- synthesizing an ideal audio data stream free of any audio quality artifacts, the ideal audio data stream comprising one or more clear utterances;
- processing the ideal audio data stream to generate another transcript, the another transcript comprising additional candidate text for each clear utterance in the ideal audio data stream along with additional confidence measures;
- and
- calculating an average confidence measure for the additional candidate text as the audio quality baseline;
- and
- compare the audio quality score to the audio quality baseline to determine whether the audio quality score indicates that one or more audio artifacts are present in the audio data stream.
13 more elements
Dependencies are inferred from the claim text.