OBFUSCATING SENSITIVE CONTENT IN AUDIO SOURCES

• US 20110218798A1
• Filed: 03/05/2010
• Published: 09/08/2011
• Est. Priority Date: 03/05/2010
• Status: Abandoned Application

• Current Assignee: Nexidia Incorporated (Nice Ltd)
• Original Assignee: Nexdia Incorporated
• Inventors: Gavalda, Marsal
• Application Number:
  US12/718,109
• Publication Number:
  US 20110218798A1
• Time in Patent Office: Days
• US Class Current: 704/201
• CPC Class Codes: G06F 16/00 Information retrieval; Database structures therefor; File system structures therefor
• CPC Class Codes: G06F 21/6245 Protecting personal data, e.g. for financial or medical purposes
• CPC Class Codes: G06F 21/6254 by anonymising data, e.g. decorrelating personal data from the owner's identification
• CPC Class Codes: G10L 15/04 Segmentation; Word boundary detection
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 17/00 Speaker identification or verification techniques
• CPC Class Codes: G10L 2015/088 Word spotting
• CPC Class Codes: G10L 21/00 Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility G10L19/00 takes precedence
• CPC Class Codes: H04M 2201/40 using speech recognition speech recognition per se G10L15/00
• CPC Class Codes: H04M 2203/6009 Personal information, e.g. profiles or personal directories being only provided to authorised persons
• CPC Class Codes: H04M 3/42221 Conversation recording systems at the subscriber's set H04M1/656

Abstract

Techniques implemented as systems, methods, and apparatuses, including computer program products, for obfuscating sensitive content in an audio source representative of an interaction between a contact center caller and a contact center agent. The techniques include performing, by an analysis engine of a contact center system, a context-sensitive content analysis of the audio source to identify each audio source segment that includes content determined by the analysis engine to be sensitive content based on its context; and processing, by an obfuscation engine of the contact center system, one or more identified audio source segments to generate corresponding altered audio source segments each including obfuscated sensitive content.

Claims

1. A method for obfuscating sensitive content in an audio source representative of an interaction between a contact center caller and a contact center agent, the method comprising:

performing, by an analysis engine of a contact center system, a context-sensitive content analysis of the audio source to identify each audio source segment that includes content determined by the analysis engine to be sensitive content based on its context; and

processing, by an obfuscation engine of the contact center system, one or more identified audio source segments to generate corresponding altered audio source segments each including obfuscated sensitive content.

2. The method of claim 1, further comprising:

preprocessing the audio source to generate a phonetic representation of the audio source.

3. The method of claim 1, wherein performing the context-sensitive content analysis includes:

searching audio data according to a search query to identify putative occurrences of the search query in the audio source, wherein the search query defines a context pattern for sensitive content; and

for each identified putative occurrence of the search query in the audio source, examining content of an audio source segment that excludes at least some portion of an audio source segment corresponding to the identified putative occurrence of the search query to determine whether linguistic units corresponding to a content pattern for sensitive content are present in the examined content.

4. The method of claim 3, wherein searching the audio data according to the search query includes determining a quantity related to a probability that the search query occurred in the audio source.

5. The method of claim 3, wherein the search query further defines the content pattern for sensitive content.

6. The method of claim 3, further comprising:

accepting the search query, wherein the accepted search query is specified using Boolean logic, the search query including terms and one or more connectors.

7. The method of claim 6, wherein at least one of the connectors specifies a time-based relationship between terms.

8. The method of claim 3, wherein the search query is accepted via a text-based interface, an audio-based interface, or some combination thereof.

9. The method of claim 3, wherein the search query is one of a plurality of predefined search strings for which the audio data is searched to identify putative occurrences of the respective search strings in the audio source.

10. The method of claim 3, wherein the search query comprises a search lattice formed by a plurality of predefined search strings for which the audio data is searched to identify putative occurrences of the respective search strings in the audio source.

11. The method of claim 1, wherein performing the context-sensitive content analysis includes:

determining a start time and an end time of each audio source segment.

12. The method of claim 11, wherein the start time, the end time, or both are determined based at least in part on one of the following:

a speaker change detection, a speaking rate detection, an elapsing of a fixed duration of time, an elapsing of a variable duration of time, a contextual pattern of content in a subsequent audio source segment, and voice activity information.

13. The method of claim 1, wherein processing one or more identified audio source segments to generate corresponding altered audio source segments includes:

substantially reducing a volume of at least a first of the one or more audio source segments to render its corresponding sensitive content inaudible.

14. The method of claim 1, wherein processing one or more identified audio source segments to generate corresponding altered audio source segments includes:

substantially masking at least a first of the one or more audio source segments to render its corresponding sensitive content unintelligible.

15. The method of claim 1, wherein processing one or more identified audio source segments to generate corresponding altered audio source segments includes:

redacting at least a portion of a first of the one or more audio source segments to render its corresponding sensitive content unintelligible.

16. The method of claim 15, wherein processing one or more identified audio source segments to generate corresponding altered audio source segments further includes:

storing the portion of the first of the one or more audio source segments that is redacted as supplemental information metadata.

17. The method of claim 1, further comprising:

permanently removing the one or more identified audio source segments prior to storing a modified version of the audio source representative of the interaction between the contact center caller and the contact center agent.

18. The method of claim 1, further comprising:

combining the altered audio source segments with unaltered segments of the audio source prior to storing a result of the combination as a modified version of the audio source representative of the interaction between the contact center caller and the contact center agent.

19. The method of claim 1, further comprising:

storing the altered audio source segments and unaltered segments of the audio source in association with a value that uniquely identifies the interaction between the contact center caller and the contact center agent.

20. The method of claim 1, wherein the sensitive content includes one or more of the following:

a credit card number, a credit card expiration date, a credit card security code, a personal identification number, and a personal authorization code.

21. The method of claim 1, wherein the context-sensitive content analysis of the audio source, the generation of corresponding altered audio source segments, or both occur substantially in real-time.

22. The method of claim 1, wherein the context-sensitive content analysis of the audio source, the generation of corresponding altered audio source segments, or both occur offline.