Real-time contact center speech analytics, with critical call alerts, deployed across multiple security zones
The invention relates to systems/methods that enable real-time monitoring/processing of contact center communications to provide timely, actionable analytic insights and real-time critical call alerts, while simultaneously providing best-in-class protection of sensitive customer information.
1. A telephonic contact center monitoring system, comprising:
a first security zone comprising:
a direct-to-transcription (DtT) adapter module that receives contact center telephony data indicative of multiple agent-caller communications, said DtT adapter module is configured to separate the received telephony data into tagged utterances, each representing a single utterance spoken by either an agent or a caller; and
a privacy-filtering automatic speech recognition (ASR) engine configured to process each utterance into a corresponding sanitized ASR transcription; and
a second security zone comprising:
a database that receives the tagged utterances that is unredacted, wherein the database is configured to update to include each tagged utterance and wherein the database is further configured to upon termination of a call, delete all utterances associated with the call except for a pending critical follow-up action.
2. The system of claim 1 , wherein the second security zone permits access by fewer users than the first security zone.
3. The system of claim 1 , wherein the first security zone has fewer security restrictions in comparison to the second security zone.
4. The system of claim 1 , wherein the second security zone hosts a critical response team that investigates critical call alert by retrieving, from the database, unredacted utterance(s) associated with the identified critical call.
5. The system of claim 4 , wherein the second security zone further includes a speech browser, configured to display the corresponding sanitized ASR transcript(s) and play the corresponding unredacted utterance(s) associated with the identified critical call.
6. The system of claim 1 , wherein the DtT adapter module operates without storing contact center telephony data in non-volatile storage locations.
7. The system of claim 1 , wherein the privacy-filtering ASR engine is further configured to remove/whitewash audio data that corresponds to the transcribed utterance, subsequent to transcription of an utterance, from any associated computer readable storage device(s).
8. The system of claim 1 , wherein the privacy-filtering ASR engine comprises (i) an ASR engine that transcribes utterances and (ii) a post-ASR redaction engine that redacts transcription in accordance with specified redaction criteria.
9. The system of claim 8 , wherein the ASR engine is stateless and the post-ASR redaction engine is stateful.
10. The system of claim 1 , wherein the privacy-filtering ASR engine comprises a privacy-by-design speech-to-text (STT) engine configured to transcribe non-sensitive information in accordance with an associated privacy-by-design language model.
11. The system of claim 1 , further comprising means for selecting class(es) of sensitive information to tokenize, including one or more of:
(1) personal names or identifying numbers;
(2) ages;
(3) locations;
(4) organizations or entities; and/or
(5) health conditions, procedures or treatments.
12. The system of claim 11 , wherein said means for selecting class(es) of sensitive information to tokenize further comprises means for selecting one or more of the selected class(es) for stratified tokenization.
13. The system of claim 12 , wherein the means for selecting comprises one of: (i) a graphical user interface (GUI), (ii) a command line interface, or (iii) an application programing interface (API).
14. The process of claim 8 , wherein the steps performed within the first security zone further include:
providing analytics based on the sanitized ASR transcriptions.
15. A telephonic contact center monitoring system, comprising:
a first security zone comprising:
a direct-to-transcription (DtT) adapter module that receives contact center telephony data indicative of multiple agent-caller communications, said DtT adapter module configured to separate the received telephony data into tagged utterances, each representing a single utterance spoken by either an agent or a caller;
an audio database, updated to include each tagged utterance;
a privacy-filtering ASR engine configured to process each utterance into a corresponding unredacted and redacted transcriptions; and
a second security zone comprising a text analytics module, configured to generate real time and post-call analytics from the redacted transcriptions.
16. The system of claim 15 further comprising a natural language processing/machine learning (NLP/ML) classifier module configured to generate an alert based, at least in part, upon one or more of the unredacted ASR transcription(s) that is sanitized and further based on one or more of the tagged utterance(s).
17. The system of claim 15 , wherein the privacy-filtering ASR Engine comprises an ASR engine to transcribe each utterance and a post-ASR redaction engine, and wherein the privacy-filtering ASR engine is further configured to redact each transcription in accordance with specified redaction criteria.
18. The system of claim 17 , wherein the ASR engine is stateless and wherein the post-ASR redaction engine is stateful.
19. The system of claim 15 , further comprising means for selecting class(es) of sensitive information to tokenize, including one or more of:
(1) personal names or identifying numbers;
(2) ages;
(3) locations;
(4) organizations or entities; and/or
(5) health conditions, procedures or treatments.
20. The system of claim 19 , wherein said means for selecting class(es) of sensitive information to tokenize further comprises means for selecting one or more of the selected class(es) for stratified tokenization.