System and method for determining the compliance of agent scripts

1. A method of script identification in audio data, the method comprising:

receiving audio data, wherein the audio data includes speech from a specific customer service agent;

performing voice activity detection to segment the audio data into a plurality of utterances, wherein each utterance is a segment of speech separated by a segment of non-speech;

performing speech analytics on each of the plurality of utterances, wherein the speech analytics includes:

filtering the plurality of utterances into a subset of utterances attributed to the specific customer service agent, and

identifying at least one acoustic feature for each of the utterances in the subset of utterances, wherein the at least one acoustic feature is used to distinguish phonemes in each of the utterances in the subset of utterances;

decoding each of the subset of the plurality of utterances to determine whether any of a plurality of script texts are contained in the utterance, wherein the decoding includes:

receiving at least one script model compilation, wherein each script model compilation includes text from a script text, a plurality of acceptable variations for the text of the script text, and at least one speaker model, wherein the at least one speaker model is specific to the specific customer service agent, further wherein the at least one speaker model is an acoustic model for the specific customer service agent,

comparing each script model compilation to the subset of the plurality of utterances using the identified acoustic features, and

determining the script model compilations present in each utterance of the subset of the plurality of utterances;

identifying each script text corresponding to each determined script model compilation;

determining compliance of the plurality of utterances, wherein determining compliance includes:

transcribing each utterance containing the script to produce an utterance transcript,

comparing the script text associated with each utterance transcript to determine a word error rate in the utterance transcript, and

determining the compliance of the utterance transcript based on the word error rate and a error threshold for the script text; and

initiating at least one remedial action if the utterance transcript is non-compliant to present on screen guidance to the specific customer service agent on a graphical display.

2. The method of claim 1 , wherein the audio data is real-time streamed audio data.

3. The method of claim 1 , wherein the remedial action is presented to the specific customer service agent in real-time.

4. The method of claim 1 , wherein the audio data is an interaction between at least the specific customer service agent and at least one customer.

5. The method of claim 1 , wherein at least one of the plurality of utterances consists of more than a single word.

6. A method of script identification in audio data, the method comprising:

receiving audio data;

performing voice activity detection to segment the audio data into a plurality of utterances, wherein each utterance is a segment of speech separated by a segment of non-speech;

performing speech analytics on each of the plurality of utterances, wherein the speech analytics includes:

filtering the plurality of utterances into a subset of utterances attributed to any customer service agents, and

identifying at least one acoustic feature for each of the utterances in the subset of utterances, wherein the at least one acoustic feature is used to distinguish phonemes in each of the utterances in the subset of utterances;

decoding each of the subset of the plurality of utterances to determine whether any of a plurality of script texts are contained in the utterance, wherein the decoding includes:

receiving at least one script model compilation, wherein each script model compilation includes text from a script text and a plurality of acceptable variations for the text of the script text,

comparing each script model compilation to the subset of the plurality of utterances using the identified acoustic features, and

determining the script model compilations present in each utterance of the subset of the plurality of utterances;

identifying each script text corresponding to each determined script model compilation;

determining compliance of the plurality of utterances with the corresponding identified script text, wherein determining compliance includes:

transcribing each utterance containing the script to produce an utterance transcript,

comparing the script text associated with each utterance transcript to determine a word error rate in the utterance transcript, and

determining the compliance of the utterance transcript based on the word error rate and an error threshold for the script text; and

initiating at least one remedial action if the utterance transcript is non-compliant to present on screen guidance to a customer service agent on a graphical display.

7. The method of claim 6 , wherein the audio data is real-time streamed audio data.

8. The method of claim 6 , wherein the audio data is an interaction between at least a customer service agent and at least one customer.

9. The method of claim 6 , wherein at least one of the plurality of utterances consists of more than a single word.

10. A method of script identification in audio data, the method comprising:

receiving audio data, wherein the audio data includes speech from a specific customer service agent;

performing voice activity detection to segment the audio data into a plurality of utterances, wherein each utterance is a segment of speech separated by a segment of non-speech;

filtering the plurality of utterances into a subset of utterances attributed to the specific customer service agent;

performing speech analytics on the subset of the plurality of utterances to identify at least one event, wherein the event is a detection of at least one keyword in the subset of the plurality of utterances, further wherein the identified event is associated with a requirement that the specific customer service agent speak at least one of a plurality of script texts in the audio data;

determining the least one script text required to be spoken by the specific customer service agent based on the identified at least one event;

receiving at least one script model compilation associated with the at least one script text, wherein each script model compilation includes text from the associated script text, a plurality of acceptable variations for the text of the associated script text, and at least one speaker model, wherein the at least one speaker model is specific to the specific customer service agent, further wherein the at least one speaker model is an acoustic model for the specific customer service agent;

determining compliance of the plurality of utterances to each script model compilation, wherein determining compliance includes:

transcribing the subset of the plurality of utterances to produce an utterance transcript,

comparing the script model compilation to the utterance transcript to determine a word error rate in the utterance transcript, and

determining the compliance of the utterance transcript based on the word error rate and an error threshold for the script model compilation; and

initiating at least one remedial action if the utterance transcript is non-compliant to present on screen guidance to the specific customer service agent on a graphical display.

11. The method of claim 10 , wherein the audio data is real-time streamed audio data.

12. The method of claim 10 , wherein the remedial action is presented to the specific customer service agent in real-time.

13. The method of claim 10 , wherein the audio data is an interaction between at least the specific customer service agent and at least one customer.

14. The method of claim 10 , wherein at least one of the plurality of utterances consists of more than a single word.

15. A method of script identification in audio data, the method comprising:

receiving audio data;

performing voice activity detection to segment the audio data into a plurality of utterances, wherein each utterance is a segment of speech separated by a segment of non-speech;

filtering the plurality of utterances into a subset of utterances attributed to any customer service agent;

performing speech analytics on the subset of the plurality of utterances to identify at least one event, wherein the event is a detection of at least one keyword in the subset of the plurality of utterances, further wherein the identified event is associated with a requirement that a customer service agent speak at least one of a plurality of script texts in the audio data;

determining the least one script text required to be spoken in the audio data based on the identified at least one event;

receiving at least one script model compilation associated with the at least one script text, wherein each script model compilation includes text from the associated script text and a plurality of acceptable variations for the text of the associated script text;

determining compliance of the plurality of utterances to each script model compilation, wherein determining compliance includes:

transcribing the subset of the plurality of utterances to produce an utterance transcript,

comparing the script model compilation to the utterance transcript to determine a word error rate in the utterance transcript, and

determining the compliance of the utterance transcript based on the word error rate and an error threshold for the script model compilation; and

initiating at least one remedial action if the utterance transcript is non-compliant to present on screen guidance to a customer service agent on a graphical display.

16. The method of claim 15 , wherein the audio data is real-time streamed audio data.

17. The method of claim 15 , wherein the audio data is an interaction between at least a customer service agent and at least one customer.

18. The method of claim 15 , wherein at least one of the plurality of utterances consists of more than a single word.