Overview
Voci's redaction feature recognizes and removes sensitive numeric values from transcripts and audio. Enable redaction by setting the values of scrubtext
(redacts transcripts) and scrubaudio
(redacts audio) to true
when submitting audio for transcription. Refer to Using Redaction for more information on enabling redaction for each Voci product.
The ASR engine is configured to redact all numbers by default with the following exceptions:
Ordinal numbers
Percentages
Times
Prices
Short decimal numbers
You can customize the information you do not want redacted by creating redaction rules. Redaction rules allow to create your own matching specifications. In general, the best practice is to redact the following:
Fifteen and sixteen digit numbers – Visa and MasterCard are sixteen digits and American Express is fifteen digits
Four and five digit numbers – People typically speak in blocks of four or five digits when providing credit card numbers
Nine digit numbers – Social Security numbers
Three digit numbers – Credit Verification Value (CVV) numbers
Note
Voci recommends the cautious approach of redacting all numbers.
The following sections explain the syntax of redaction rules, how to create them, and how to use them with V‑Blaze, V‑Cloud, and V‑Spark.