Secure call transcriptions allow the system to detect when credit card information, social security information, or phone numbers are spoken during a call, tag the call appropriately, and redact that information from your call transcriptions and associated call recording.
These optional security features are available for basic transcriptions only. If any security features are enabled, enhanced transcriptions cannot be used.
- PCI Detection detects when a credit card number or CCV is spoken during a call. If either is detected, a “PCI” tag will be applied to the call. More information on PCI can be found here: PCI Compliance
- SSN Detection detects when a Social Security Number is spoken during a phone call. If detected, an “SSN” tag will be applied to the call.
- Number Detection detects when a number is spoken during a call. If detected, a “number” tag will be applied to the call.
- Redaction will automatically remove any sensitive information detected in the transcription. If this is done, a “redacted” tag will be applied to the call.
Enabling Secure Transcriptions
- Navigate to Numbers → Call Settings.
- Click edit next to the configuration you wish to change.
- Click or scroll down to the Transcriptions section.
- If you have not already done so, click to enable transcriptions for this call configuration. Enhanced Transcription does not support redaction or tagging at this time.
- Once transcriptions are enabled, new options will appear. Click the toggles to select what types of information you would like to have detected. For any option you have selected, calls will be tagged whenever that type of information is recognized in the recording of a call.
- If you wish to redact any of the selected information types from your transcriptions and recordings, click to enable redaction. With redaction enabled, protected information will be covered with a low tone for that section of the call recording, and the corresponding transcription will be replaced with [redacted] text.
- Click Save Changes.
PCI detection and redaction uses a training set that includes examples of PCI dialogs from a number of English-speaking countries. We are continually adding new features and algorithms to improve PCI detection and reduce redaction in recordings that do not contain PCI information. The complexity of PCI detection is a result of factors such as customer repetition, agent readback, declined cards, inaudible speech, and poor call quality.
Summary of Latest Test Results:
- 99% of files containing PCI were identified and effectively redacted
- 3% of recordings that did not contain PCI had modest unnecessary redaction (5-10%)
While we do not guarantee or commit to specific accuracy rates, the following are reasonable expectations:
- More than 2/3 of numbers redacted in most recordings
- Redaction in >> 90% of recordings – “in the high 90’s”
- Modest amounts of excess redaction, almost none in non-PCI recordings